box_process_call() uses func_call(), which not only calls the given
function, but also checks that the current user has the right to execute
it. As a result, we can't add auditing for only those function calls
that passed the access check (apparently, there's no reason to log
function calls that failed with an 'access denied' error - we have a
separate audit event for this).

To fix this, let's introduce func_call_no_access_check() helper, which
calls a function without checking access rights, and use it along with
existing func_access_check() in box_process_call(). func_call() is now
an inline function that calls func_access_check() and then on success
func_call_no_access_check().

It's probably wrong that func_call() checks access rights, because this
means that to use a space with a functional index/constraint, the user
needs not only read/write access to the space itself, but also execute
access to the function. I think we should check the right to execute
such function only once - on functional index/constraint creation, not
on every call, but I'm not going to change this now, because nobody's
complained so far, and a change like this needs a proper discussion
anyway.

NO_TEST=refactoring
NO_DOC=refactoring
NO_CHANGELOG=refactoring

The trigger is invoked on console eval. The trigger callback is passed
the eval expression string. It will be used for auditing console events
in the EE version.

NO_TEST=ee
NO_DOC=internal
NO_CHANGELOG=internal

Simply add Lua wrappers for recently introduced prbuf.

Introduce new buffer (in addition to ibuf) - prbuf. "pr" stands for
"partitioned ring-". It save all metadata in the same memory chunk
provided for storage, so it can be completely restored from the 'raw'
memory. API:

```
-- mem is a chunk of raw (char *) memory, of size mem_size.
-- It is used for data storage. Note that available space is of less
-- size due to prbuf metadata overhead.
-- Returns handle to prbuf.
--
require('buffer').prbuf_create(mem, mem_size)

-- mem is a chunk of memory, which contains already created prbuf.
-- It implies that once prbuf_create() was called with the same memory.
-- If mem does not contain valid buffer - raise an appropriate error.
require('buffer').prbuf_open(mem)

-- Returns continuous chunk of memory with given size. May return nil
-- in case if requested chunk is too large. Note that after copying
-- object to returned chunk, it should be committed with prbuf:commit();
-- otherwise :prepare() may return the same chunk twice.
prbuf:prepare(size)

-- Commits the last prepared chunk. Undefined behaviour in case of
-- committing the same chunk twice.
prbuf:commit()

-- Create and return prbuf_iterator. Does not fail. Created iterator
-- points to nowhere - it should be adjusted with :next() call to
-- the first entry.
prbuf:iterator_create()

-- Advance iterator position. Returns prbuf_entry or nil in case
-- iterator has reached the end. Entry consists of two members:
-- size and ptr. The last one is an array of characters of given size.
iterator:next()
```

 Usage examples:

```

local ibuf = buffer.ibuf()
local prbuf_size = 100
local memory = ibuf:alloc(prbuf_size)
local prbuf = buffer.prbuf_create(memory, prbuf_size)

local sample_size = 4
local raw = prbuf:prepare(4)
if raw == nil then
    -- allocation size is too large, try smaller.
end
raw[0] = ...
...
raw[3] = ...
prbuf:commit()
local prbuf_recovered = buffer.prbuf_open(memory)
local iter = prbuf_recovered:iterator_create()
local entry = iter:next()
assert(tonumber(entry.size) == 4)
-- Check values stored in the buffer.
assert(entry.ptr[0] == ...)
entry = iter:next()
-- Our buffer has only one entry.
assert(entry == nil)

```

NO_DOC=<Feature for internal usage>
NO_CHANGELOG=<Feature for internal usage>

Required for backward compatibility.

Follow-up commit a303b53b ("cfg: add audit_format option").

NO_DOC=ee
NO_CHANGELOG=ee

prbuf is partitioned ring buffer. The main property of the buffer is
that it can be recovered from raw memory. To achieve this buffer saves
metadata before each stored entry. For further details see source code.

NO_DOC=<Private data structure>
NO_CHANGELOG=<No user visible changes>

This patch introduces memtx_tx_region and memtx_tx_mempool:
engineers must use only these proxies to collect statistics.

Also this patch introduces box.stat.memtx.mvcc - the way to
get memtx mvcc memory statistics.

Closes #6150

@TarantoolBot document
Title: Memtx MVCC memory monitoring

Introduce memtx MVCC memory monitoring. One can get it with
box.stat.memtx.tx() method or use index to access a particular
statistic. The statistics format:

txn:
  statements:
    max: 0
    avg: 0
    total: 0
  user:
    max: 0
    avg: 0
    total: 0
  system:
    max: 0
    avg: 0
    total: 0
mvcc:
  trackers:
    max: 0
    avg: 0
    total: 0
  conflicts:
    max: 0
    avg: 0
    total: 0
  tuples:
    tracking:
      stories:
        count: 0
        total: 0
      retained:
        count: 0
        total: 0
    used:
      stories:
        count: 0
        total: 0
      retained:
        count: 0
        total: 0
    read_view:
      stories:
        count: 0
        total: 0
      retained:
        count: 0
        total: 0

Since transaction can allocate objects on its own, we need to track it
to collect memory statistics of memtx txm. Txn's use only region
allocator so this patch introduces tx_region methods - txns should use
it instead of region methods to collect allocation statistics.

Part of #6150

NO_DOC=no visible changes
NO_CHANGELOG=no visible changes
NO_TEST=already covered

Since a transaction has been prepared its garbage (produced stories
and replaced tuples) cannot be deleted because they are recognized
as used in read-view even if there are no transactions in read-view.
This problem makes it difficult to test the memory monitoring system,
so this patch solves the problem.

Close #6635
Part of #6150

NO_DOC=no visible changes
NO_CHANGELOG=no visible changes

NO_DOC=refactoring
NO_CHANGELOG=refactoring
NO_TEST=refactoring

It is difficult to monitor memory consumption for each transcation
when some functions allocating on txn's region do not have any information
about owner of the region. This patch solve the problem.

Part of #6150

NO_DOC=refactoring
NO_CHANGELOG=refactoring
NO_TEST=refactoring

An opportunity to call garbage collector of memtx transaction
manager manually allows to understand which garbage cannot be freed.
This knowledge can help us to improve garbage collector. Also this
opportunity makes it easier to test memtx mvcc memory monitoring.

Part of #6150

NO_DOC=internal feature
NO_CHANGELOG=internal feature
NO_TEST=internal feature

There is no information about owner of txn_stmt when it is
being allocated. It makes difficult to track memory
consumption for each transaction. That is why we need to
pass transaction itself to txn_stmt_new() instead of its region.

Part of #6150

NO_DOC=refactoring
NO_CHANGELOG=refactoring
NO_TEST=refactoring

Commit/rollback triggers are run asynchronously, upon receiving the
write status from WAL. We can't run them in the original fiber that
submitted the WAL request, because it would open a time window between
writing a transaction to WAL and committing it in tx, which could lead
to violating the cascading rolback principles. As a result,
commit/rollback triggers run with admin privileges.

Let's fix this issue by temporarily setting session and credentials to
the original fiber's for running commit/rollback triggers.

Closes #7005

NO_DOC=bugfix

Tarantool used to crash if launched with multiple -e or -l options without
a space between the option and the value, e.g.: `tarantool -ea -eb`.
It happened because optv[] was allocated for argc==3 items, while 4
options were written to it after parsing (-e, a, -e, b).
This patch allocates optv[] for the maximum possible number of options:
(argc - 1) * 2.

Closes #5747

NO_DOC=bugfix

The option will be used in the Enterprise version to configure audit
log event filter.

NO_DOC=ee
NO_CHANGELOG=ee

The functions are not used since commit 1a6ad79e ("lua/cfg: drop
unused log methods").

NO_DOC=drop unused code
NO_TEST=drop unused code
NO_CHANGELOG=drop unused code

This patch fixes an issue with the implicit cast of INTEGER and DECIMAL
values to DOUBLE when they are passed as the first argument to the
ROUND() function.

Closes #6988

@TarantoolBot document
Title: ROUND() now properly supports INTEGER and DECIMAL

INTEGER and DECIMAL values passed as the first argument now will not be
cast to DOUBLE and the result will be of the same type as the first
argument. Also, the default type for the ROUND() is now DECIMAL.

The option will be used in the Enterprise version to configure audit
log format.

NO_DOC=ee
NO_CHANGELOG=ee

The function doesn't modify neither the array passed to it, nor the
strings in the array so it should use const.

NO_DOC=refactoring
NO_TEST=refactoring
NO_CHANGELOG=refactoring

Introduce new option IPROTO_TXN_ISOLATION (0x59) in the body of
IPROTO_BEGIN request, so a user can set isolation level similar
to box.begin in lua.

The value must be one of the following integers:
enum txn_isolation_level {
	/** Take isolation level from global default level. */
	TXN_ISOLATION_DEFAULT,
	/** Allow to read committed, but not confirmed changes. */
	TXN_ISOLATION_READ_COMMITTED,
	/** Allow to read only confirmed changes. */
	TXN_ISOLATION_READ_CONFIRMED,
	/** Determine isolation level automatically. */
	TXN_ISOLATION_BEST_EFFORT,
};

Support the new option in net.box.

Part of #6930
NO_DOC=see later commits
NO_CHANGELOG=see later commits

Now memtx TX manager tries to determine the best isolation level
by itself. There could be two options:
* READ_COMMITTED, when the transaction see changes of other tx
that are committed but not yet confirmed (written to WAL)
* READ_CONFIRMED, when the transaction see only confirmed changes.

Introduce a simple way to specify the isolation level explicitly:
box.begin{tx_isolation = 'default'} - the same as box.begin().
box.begin{tx_isolation = 'read-committed'} - READ_COMMITTED.
box.begin{tx_isolation = 'read-confirmed'} - READ_CONFIRMED.
box.begin{tx_isolation = 'best-effort'} - old automatic way.

Intrduce a bit more complex but potentially faster way to set
isolation level, like that:
my_level = box.tnx_isolation_level.READ_COMMITTED
..
box.begin{tx_isolation = my_level}

For simplicity of implementation also support symmetric values as
'READ_COMMITTED' and box.tnx_isolation_level['read-committed'].

Introduce a new box.cfg option - default_tx_isolation, that is
used as a default when a transaction is started. The option is
dynamic and possible values are the same as in box.begin, except
'default' which is meaningless.

In addition to string value the corresponding numeric values can
be used in both box.begin and box.cfg.

Part of #6930
NO_DOC=see later commits
NO_CHANGELOG=see later commits

When a transaction is started without specifying isolation level
(which is impossible now) the transactional manager must choose
the transaction level automatically, that means that is must
detemine whether the transaction can see other prepared changes
or not. The best effort that we can made is to check if current
transaction is read-only or not. For read-only transactions
there are hope and fear that it will remain read-only, and the
best choice is not to see prepared changes. But if the transaction
has DML statements - it must see prepared changed.

Note that a read-only transaction can became read-write if it make
a DML statement. But if a transaction ignores some other prepared
change and then makes a DML, there are no other options except
abort that transaction - it could not be serialized anymore.

Part of #6930
Closes #6246
NO_DOC=see later commits

This flag actually describes isolation level:
is_prepared_ok == true - READ_COMMITTED,
is_prepared_ok == false - READ_CONFIRMED.

Now it is always calculated as tnx != NULL.
Let memtx_tx_tuple_clarify to calculate it by itself.

No logical changes.

Part of #6930
NO_DOC=refactoring
NO_CHANGELOG=refactoring
NO_TEST=refactoring

No logical changes.

Part of #6930
NO_DOC=refactoring
NO_CHANGELOG=refactoring
NO_TEST=refactoring

Internally space's indexes can containt dirty tuples that are
invisible for user. That's why there's special adjustment in
space:count() that substracts number of invisible tuple in the
space.

By a mistake that check thought that all prepared statements are
visible, which is wrong for autocommit reads, like standalone
space:count() without explicit transaction.

Fix it by using common for all reads practice: ignore prepared
statements if current transaction is NULL.

Closes #6421
NO_DOC=bugfix

Memtx TX manager stores a pointer to deleting statement in prepared
story. This pointer is set in two cases:
1. a statement deletes (or overwrites) a tuple
2. a story becomes prepared while other inprogress TX overwrites it

By design a tuple can be deleted only by primary index, the case
when a transaction overwrites somehting in secondary index but does
not overwrite the same tuple in primary index is prohibited. That's
why the pointer (to deleting statement) must be set by and only by
the next statement in the primary index chain.

By mistake the pointer is set also in second index chain analysis
after reordering which led to unexpected state of a story.

The patch removes the problem.

Closes #6452
NO_DOC=bugfix

We used to use very ugly and tricky approach to check that passed years,
months and days were not exceeding supported range of values. Now we have
introduced to `c-dt` library the new function `dt_from_ymd_checked` for
that purpose (i.e. check that values are valid, and construct dt from
them). So rewrite/simplify Lua code to use that entry as
`tnt_dt_from_ymd_checked`.

Part of #6731

NO_DOC=refactoring
NO_CHANGELOG=refactoring

* Default parse
  - new c-dt version used which handles extended years range
    while parse relaxed iso8601 gformat strings;
  - family of functions like dt_from_ymd_checked functions
    added to the new c-dt version, now used by conversion code
    to properly handle validation of a 32-bit boundary values;
  - datetime_parse_full() modified to properly handle huge years values;
  - added tests for extended years range.

* strptime-like parse
  - properly handle longer than 4 years values, negative values,
    and handle zulu suffix, which may be generated by Tarantool
    stringization routines;

Part of #6731

NO_DOC=internal
NO_CHANGELOG=internal

To parse date/time strings using format string we use
`strptime()` implementation from FreeBSD, which is
modified to use our `struct datetime` data structure.

List of supported format has been extended to include
`%f` which is flag used whenever you need to process
nanoseconds part of datetime value.

```
tarantool> T = date.parse('Thu Jan  1 03:00:00 1970', {format = '%c'})

tarantool> T
- 1970-01-01T03:00:00Z

tarantool> T = date.parse('12/31/2020', {format = '%m/%d/%y'})

tarantool> T
- 2020-12-31T00:00:00Z

tarantool> T = date.parse('1970-01-01T03:00:00.125000000+0300',
                          {format = '%FT%T.%f%z'})

tarantool> T
- 1970-01-01T03:00:00.125+0300
```

Part of #6731

NO_DOC=internal
NO_CHANGELOG=internal

Datetime module provides parse function to create
datetime object given input string.

`datetime.parse` function expect 1 required argument - which is
input string, and set of optional parameters passed as table
in 2nd argument.

Allowed attributes in this optional table are:
* `format` - should be either 'iso8601', 'rfc3339' or `strptime`-like
  format string. [strptime format will be added as part of next
  commit];
* `tzoffset` - to redefine offset of input string value, if there
  is no timezone provided.
* `tz` - human-readable, Olson database, timezone identifier, e.g.
  'Europe/Moscow'. Not yet implemented in this commit.

```
tarantool> T = date.parse('1970-01-01T00:00:00Z')

tarantool> T
- 1970-01-01T00:00:00Z

tarantool> T = date.parse('1970-01-01T00:00:00',
                          {format = 'iso8601', tzoffset = 180})

tarantool> T
- 1970-01-01T00:00:00+0300

tarantool> T = date.parse('2017-12-27T18:45:32.999999-05:00',
                          {format = 'rfc3339'})

tarantool> T
- 2017-12-27T18:45:32.999999-0500
```

Implemented as per RFC https://hackmd.io/@Mons/S1Vfc_axK#%D0%AD%D1%82%D0%B0%D0%BF-3-%D0%9F%D0%B0%D1%80%D1%81%D0%B8%D0%BD%D0%B3-%D0%B4%D0%B0%D1%82-%D0%BF%D0%BE-%D1%84%D0%BE%D1%80%D0%BC%D0%B0%D1%82%D1%83

Part of #6731

NO_DOC=internal
NO_CHANGELOG=internal

When the master is just starting up it's possible for replica's JOIN
request to arrive right in time to bypass ER_LOADING check (after master
is fully recovered) but still fail due to ER_READONLY: box.cfg.read_only
is only read and set after box_cfg() (its C part) returns.

In this case the joining replica simply exits with an error and doesn't
retry JOIN.

Let's fix that. Make ER_READONLY a recoverable error and let replica
retry joining after receiving ER_READONLY.

Anonymous nodes relied on ER_READONLY to forbid replication from
anonymous to normal replicas. That check doesn't work anymore.
So introduce explicit checks banning replication from anonymous nodes.

Note, there were some alternatives to this fix.

First of all, theoretically, we could stop firing ER_LOADING later,
after box_cfg() is complete. This solution wouldn't work because it
would lead to deadlocks: the nodes would be stuck in replicaset_sync(),
because each of them rejects replication with ER_LOADING.

Another solution would be to read the real box.cfg.read_only value
earlier, in order to allow replication right after the node finishes
recovery.
This would also be bad, because we should never let a node become
writeable before box.cfg is finished. Even after local_recovery is
complete, the node should stay read-only until it synchronizes with
other replicas.

That said, neither of the two alternatives fit, so the solution with
retrying JOIN on ER_READONLY was chosen.

Since the bug is fixed, re-enable the test in which it was discovered:
replication-py/init_storage.test.py

Also, remove replication/ddl.test.lua from fragile list, since this bug
was the only reason for its fragility.

Closes #5337
Closes #6966

NO_DOC=minor bugfix

The source file will be added to the Enterprise repository.

NO_DOC=cmake
NO_TEST=cmake
NO_CHANGELOG=cmake

fiber_new_ex() used to ignore fiber_attr flags when the fiber was
taken from the cache, not created anew.

It didn't matter much though for the public API, because the only
public flag in fiber_attr was FIBER_CUSTOM_STACK (which can be
set via fiber_attr_setstacksize()).

Anyway that was a bug for internal API and would lead to issues in
the future when more public flags are added. The patch fixes it.

NO_DOC=Bugfix
NO_CHANGELOG=No reproducer via public API

There was a user who complained about this code crashing:

    f = fiber_new_ex(...);
    fiber_start(f);
    fiber_cancel(f);

The crash was at cancel. It happened because the fiber finished
immediately. It was already recycled after fiber_start() return.

Recycled fiber didn't have any flags, so fiber_cancel() didn't
see the fiber was already dead and tried to wake it up. It crashed
when the fiber tried to call its 'fiber->f' function which was
NULL.

In debug build the process fails earlier with an assertion on
'fiber->fid != 0'.

It can't be really fixed because the problem is the same as with
use-after-free. The fiber could be not recycled but already freed
completely, returned back to the mempool.

This patch tries to help the users by a panic with a message
saying that it wasn't just a crash, it is a bug in user's code.

There is an alternative - make fibers never return to the mempool.
Then fiber_cancel() could ignore recycled fibers. But it would
lead to another problem that if the fiber is already reused, then
fiber_cancel() would hit a totally irrelevant fiber who was
unlucky to reuse that fiber pointer. It seems worse than panic.

Same problem exists for `fiber_wakeup()`, but I couldn't figure
out how to add a panic there and not add an `if` on the normal
execution path (which includes 'ready' and 'running' fibers).

Closes #6837

NO_CHANGELOG=The same crash remains, but happens a bit earlier and
  with a message.

@TarantoolBot document
Title: `fiber_cancel()` C API clarification

The documentation must warn that the fiber passed to
`fiber_cancel()` must not be already dead unless it was set to be
joinable. Same for `fiber_wakeup()` and all the other fiber
functions. A dead non-joinable fiber could already be freed or
reused.

Fibers with custom stack couldn't be reused via cord->dead list,
but neither were ever deleted via mempool_free(). They just leaked
until the cord was destroyed. Their custom stack also leaked.

It happened for all non-joinable custom-stack fibers. That was
because fiber_destroy() simply skipped the destruction if the
fiber is the current one.

It didn't affect joinable fibers because their fiber_destroy() is
done in another fiber. Their stack was deleted, but the fiber
itself still leaked.

The fix makes so fiber_destroy() is never called for the current
fiber. Instead, cord uses an approach like in pthread library -
the fiber who wants to be deleted is saved into cord->garbage
member. When some other fiber will want to be deleted in the
future, it will firstly cleanup the previous one and put self into
its place. And so on - fibers cleanup each other.

The process is optimized for the case when the fiber to delete is
not the current one - can delete it right away then.

NO_DOC=Bugfix

As a result of recording <crc32:update> method or <digest.crc32>
function wrong semantics is compiled (strictly saying, the resulting
trace produces the different result from the one yielded by
interpreter). The easiest solution is disabling JIT for particular
functions, however, such approach drops the overall platform
performance. Hence, the mentioned functions are rewritten line by line
via Lua C API to avoid JIT misbehaviour.

NO_DOC=no visible changes
NO_CHANGELOG=no visible changes

During the context switch required for backtracing a suspended fiber,
unwinders go crazy, as the unwind information they had gets implicitly
invalidated: provide an annotation for a dummy frame for
`coro_unwcontext`, as if it were at the bottom of the call-chain — that
way unwinders can normally proceed further.

We need to know the exact location of the stack pointer: replace the
16-byte stack alignment instruction on x86_64 macOS by adding the
`force_align_arg_pointer` attribute to `unw_getcontext_f`.

Needed for #4002

NO_DOC=bug fix
NO_CHANGELOG=bug fix
NO_TEST=unwind information annotation in inline assembly

NO_DOC=no behavior changes
NO_TEST=no behavior changes
NO_CHANGELOG=no behavior changes

Added predefined system events: box.status, box.id, box.election and
box.schema.

Closes #6260

@TarantoolBot document
Title: Built-in events for pub/sub

Built-in events are needed, first of all, in order to learn who is the
master, unless it is defined in an application specific way. Knowing who
is the master is necessary to send changes to a correct instance, and
probably make reads of the most actual data if it is important. Also
defined more built-in events for other mutable properties like leader
state change, his election role and election term, schema version change
and instance state.

Built-in events have a special naming schema - their name always starts
with box.. The prefix is reserved for built-in events. Creating new events
with this prefix is banned. Below is a list of all the events + their names
and values:

1. box.id
Description - identification of the instance. Changes are extra rare. Some
values never change or change only once. For example, instance UUID never
changes after the first box.cfg. But is not known before box.cfg is called.
Replicaset UUID is unknown until the instance joins to a replicaset or
bootsa new one, but the events are supposed to start working before that -
right at listen launch. Instance numeric ID is known only after
registration. On anonymous replicas is 0 until they are registered
officially.
Value - {
    MP_STR “id”: MP_UINT; box.info.id,
    MP_STR “instance_uuid”: MP_UUID; box.info.uuid,
    MP_STR “replicaset_uuid”: MP_UUID box.info.cluster.uuid,
}

2. box.status
Description - generic blob about instance status. Its most commonly used
and not frequently changed config options and box.info fields.
Value - {
    MP_STR “is_ro”: MP_BOOL box.info.ro,
    MP_STR “is_ro_cfg”: MP_BOOL box.cfg.read_only,
    MP_STR “status”: MP_STR box.info.status,
}

3. box.election
Description - all the needed parts of box.info.election needed to find who
is the most recent writable leader.
Value - {
    MP_STR “term”: MP_UINT box.info.election.term,
    MP_STR “role”: MP_STR box.info.election.state,
    MP_STR “is_ro”: MP_BOOL box.info.ro,
    MP_STR “leader”: MP_UINT box.info.election.leader,
}

4. box.schema
Description - schema-related data. Currently it is only version.
Value - {
    MP_STR “version”: MP_UINT schema_version,
}

Built-in events can't be override. Meaning, users can't be able to call
box.broadcast(‘box.id’, any_data) etc.

The events are available from the very beginning as not MP_NIL. It's
necessary for supported local subscriptions. Otherwise no way to detect
whether an event is even supported at all by this Tarantool version. If
events are broadcast before box.cfg{}, then the following values will
available:
    box.id = {}
    box.schema = {}
    box.status = {}
    box.election = {}

This way the users will be able to distinguish an event being not supported
at all from box.cfg{} being not called yet. Otherwise they would need to
parse _TARANTOOL version string locally and peer_version in netbox.

Example usage:

 * Client:
   ```lua
   conn = net.box.connect(URI)
   -- Subscribe to updates of key 'box.id'
   w = conn:watch('box.id', function(key, value)
       assert(key == 'box.id')
       -- do something with value
   end)
   -- or to updates of key 'box.status'
   w = conn:watch('box.status', function(key, value)
       assert(key == 'box.status')
       -- do something with value
   end)
   -- or to updates of key 'box.election'
   w = conn:watch('box.election', function(key, value)
       assert(key == 'box.election')
       -- do something with value
   end)
   -- or to updates of key 'box.schema'
   w = conn:watch('box.schema', function(key, value)
       assert(key == 'box.schema')
       -- do something with value
   end)
   -- Unregister the watcher when it's no longer needed.
   w:unregister()
   ```

We use a special, less efficient index vtab if a space can store
compressed tuples. The problem is it isn't enough to look at a space
definition to figure out if there are compressed tuples in the space:
there may be compressed tuples left from before the alter operation that
disabled compression, since we don't rebuild tuples on alter. To update
an index vtab dynamically, we implement some complicated logic, but
it's buggy (results in a test failure in EE). Fixing it requires some
non-trivial effort, because a vtab may be changed after index creation
(when a space format is updated).

Let's drop this optimization altogether for now and use the same vtab
for both compressed and uncompressed indexes. We might return to this
issue in future, but first we need to run some benchmarks to check if
this optimization is worth the complexity. Possible ways how we could
resurrect this optimization:
 - Call get_raw from get directly (without function pointer), inline
   memtx_prepare_result_tuple, and move is_compressed flag to struct
   tuple for better cache locality.
 - Rebuild all tuples on space alter and use a different vtab for
   compressed indexes.

NO_DOC=bug fix
NO_TEST=enterprise
NO_CHANGELOG=unrelased