This patch enables static and dynamic type check for aggregate functions
SUM(), AVG() and TOTAL().

Part of #6105

Prior to this, the patch functions MIN(), MAX(), LEAST() and GREATEST()
showed SCALAR as result types in metadata. However, in reality, the type
of the result could be any scalar type. After this patch, the type of
the result will always be the same as the type in the metadata. Also,
for MIN() and MAX() functions, the type of the result will be the same
as the type of the argument. For the LEAST() and GREATEST() functions,
the result type will be the same as the type of the arguments if all
arguments are of the same type, or it will be NUMBER if all arguments
are of numeric types, or it will be SCALAR.

Part of #6105

This patch enables static and dynamic type checks for functions that do
not need any rework.

Part of #6105

This patch introduces a runtime type checking mechanism for SQL built-in
functions. However, it is currently disabled as the functions themselves
need to be prepared for such changes.

Part of #6105

This patch introduces a static type checking mechanism for SQL built-in
functions. However, it is currently disabled as the functions themselves
need to be prepared for such changes.

Part of #6105

Prior to this patch, the number of arguments for functions with a
variable number of arguments was checked at runtime. After this patch,
it will be checked during parsing. For functions with a constant number
of arguments, it is always checked during parsing.

Part of #6105

Currently, the SQL built-in hash table contains one already defined
implementation for each SQL built-in function. This is rather
inconvenient because some built-in SQL functions can accept arguments of
more than one type, and the type of the result can depend on the types
of the arguments. In addition, the number of arguments can be variable
for some built-in SQL functions. For these reasons, we are forced to
check the number of arguments and their type at runtime. To make it
possible to check types of arguments and their number during parsing,
the hash table has been modified so that functions can now have more
than one implementation.

Part of #6105

This patch changes the signature of SQL built-in function TRIM(). This
gives us an easier way to check the types of the arguments to this
function. Additionally, these changes fix a bug where using TRIM with
the BOTH, LEADING, or TRAILING keywords would result in a loss of a
collation.

Needed for #6105
Closes #6299

Follow-up #6185

After this patch SQL built-in function QUOTE() will return the argument
in case DOUBLE value is given. If the argument is not number, string
representation of the argument will be returned.

Closes #6239

@TarantoolBot document
Title: QUOTE() and DOUBLE argument

After this patch function QUOTE() will return argument in case it
receives DOUBLE value as an argument. The same for all other numeric
types. In case it was given value of non-numeric type, it will return
string representation of the argument.

New release policy defines pre-release tags: *-alpha*, *-beta*, *-rc*.
Packages with these tags must be published to pre-release folder in
the S3 repo. Was added PRERELEASE_REPO_S3_DIR path to all workflows with
packaging *.rpm and *.deb and fixed conditions for updating the S3 repo.

Needs for: #6185

New release policy provides new rules for publishing packages.
Now, only major version is defining bucket repo, for more convenience
use next naming series-<MAJOR_VERSION>.

Needs for: #6185

According to the new release policy live packages are not allowed.
Only releases and pre-releases would be published to the S3.

Needs for: #6185

This patch enables all tests in tarantool/luajit to be run in CI, since
all issues found with them are finally resolved in scope of the commit
847b096e ('luajit: bump new version').

Follows up #6098
Follows up tarantool/tarantool-qa#120

Reviewed-by: Sergey Kaplun <skaplun@tarantool.org>
Reviewed-by: Vitaliia Ioffe <v.ioffe@tarantool.org>
Signed-off-by: Igor Munkin <imun@tarantool.org>

* ARM64: Fix exit stub patching.
* arm64: fix cur_L restoration on error throw

Closes #6098
Closes #6189
Part of #5629
Relates to #6323
Follows up #1516

This patch adds changelog entry for commit
027775ff ('lua: refactor
port_lua_do_dump and encode_lua_call').

Follows up #6248
Follows up #4617

Signed-off-by: Igor Munkin <imun@tarantool.org>

This patch adds changelog entry for commit
ba5398c7 ('luajit: bump new version').

Follows up #6224
Follows up #6227

Signed-off-by: Igor Munkin <imun@tarantool.org>

This patch introduces arithmetic for DECIMAL in SQL. After this patch,
DECIMAL values can participate in arithmetic along with INTEGER,
UNSIGNED, DOUBLE, and other DECIMAL values.

Part of #4415

This patch introduces implicit and explicit cast of the DECIMAL field in
SQL. Implicit cast to and from DECIMAL is precise. Explicit conversion
to decimal and back could lead to loss of precision.

Part of #4415

This patch introduces a decimal field type. However, implicit and
explicit casts and arithmetic operations for this type will be presented
in next few patches. Literals also will be introduced later.

Part of #4415

This patch introduces function decimal_is_neg() which checks that
decimal is less than zero.

Needed for #4415

RW transactions used to commit without problems after rollback
of transactions holding stories responsible to track their reads.
That behaviour led to unserializable results.
This patch fixes this problem.

Closes #6325

There are two cases in lua trigger implementation, which leads
to access to freed memory: if trigger removes itself and if
trigger yields and someone destroy it. These problems was fixed
by remembering the necessary fields from the trigger structure
in local variables before calling trigger function.

Closes #6266

After this patch, SCALAR values will be able to be compared with values
of any other scalar type. The comparison will be done according to the
SCALAR rules, which means boolean values < numeric values < string
values < binary values < uuid values.

Closes #6221

@TarantoolBot document
Title: SCALAR and NUMBER values in SQL

SCALAR values can now be compared with values of any other scalar type,
but cannot be implicitly cast to any other scalar type. This means that
SCALAR values cannot participate in arithmetic, bitwise operations,
concatenation, or functions that, by definition, do not accept SCALAR
values.

NUMBER values now also cannot be implicitly cast to any other numeric
type, which means that NUMBER values cannot participate in arithmetic
and bitwise operations, or in functions that, by definition, do not
accept NUMBER values.

Part of #6221

Part of #6221

This patch disallows arithmetic operations for SCALAR and NUMBER values.
It also corrects the description of the error that is generated when an
inappropriate value participates in an arithmetic operation.

Part of #6221

This patch disallows implicit cast during INSERT and UPDATE from SCALAR
to any other scalar types and from NUMBER to any other numeric types.

Part of #6221

This patch re-introduces the NUMBER and SCALAR meta-types. They were
introduced some time ago, however there were some problems with how they
worked and some properties of these meta-types were not implemented.

New properties of meta-type SCALAR in SQL:
1) typeof() with SCALAR argument returns "scalar";
2) Any value of any scalar type can be implicitly cast to SCALAR;
3) SCALAR values can be compared with any value of any other scalar
   type;
4) SCALAR values cannot be implicitly cast to any other scalar type;
5) SCALAR values cannot participate in arithmetic or bitwise
   operations;
6) SCALAR values cannot be concatenated with other values.

New properties of meta-type NUMBER in SQL:
1) typeof() with NUMBER argument returns "number";
2) Any value of a numeric type can be implicitly cast to NUMBER;
3) NUMBER values can be implicitly cast to SCALAR;
4) NUMBER cannot be implicitly cast to any other numeric type;
5) NUMBER values cannot participate in arithmetic or bitwise
   operations;

This patch only applies 1) and 2) for both meta-types. All other
properties will be presented in the next patches.

Part of #6221

This patch removes the enum field_type from the struct MEM since the
only place where this field was used is in the typeof() function. After
this patch, typeof() will return "NULL" if it receives NULL. Prior to
this patch, the field type was returned if it was set, otherwise it was
returned "boolean". Another reason for the removal was that field_type
was quite volatile and could change in unexpected places.

Also, this patch made another change: values of the field type NUMBER
will now be processed by the typeof() as INTEGER or DOUBLE, depending
on the storage type. However, this will be fixed in the next patch.

Part of #5956
Needed for #6221

Prior to this patch, most ephemeral space fields were defined using the
SCALAR type. After this patch, almost all fields will be properly
defined. However, there are still cases where SCALAR will be set by
default. For example, in IN, where rules is still not defined (#4692).
Another example is when a field is not resolved because of a too complex
query.

Part of #6213

This patch renames sql_ephemeral_space_new() to sql_template_space_new()
and sql_ephemeral_space_def_new() to sql_template_space_def_new(). The
objects created by these functions are not really ephemeral space or
space_def and have a single purpose - to create a "template" space that
will be used later when creating real space.

Needed for #6213

Import of `table.new` module was removed as a part of net.box
refactoring in commit 954194a1.
The `table.new` method became unavailable in Tarantool.

This commit returns it back and adds a corresponding regression test.

Closes #6330

Fix option argument in user.grant function

It makes sense to abort only in-progress transactions while we are going
to abort all transactions on DDL TX preparation step. Already prepared,
committed or aborted transactions turn out to be uninteresting for us.

Follow-up #5998

FORCE_CONFIG=true is preventing local .rocks/config-5.1.lua from loading.
Local config is often used to populate rocks_servers with own rocks servers and with original luarocks

This patch changes FORCE_CONFIG to true in generated luarocks.core.hardcoded to allow using
project-side config-5.1.lua for luarocks configuration to avoid dirty and ugly hacks

fiber.top testing used to be broken under arm platforms because of the
assumption that fiber.top functionality must be enabled at build time.
This patch modifies the test for this case.

Follow-up #6272

MacOS has its own libunwind which works not good with fibers
under M1. Backtraces need to be enabled again after #6060 is fixed.
Other arm backtraces are temporarily disabled too for #6222.

Closes #6272

Currently, when calling 'close' method for connection,
unsent requests are lost. Disable appropriate check until
corresponding (#6338) issue will be resolved.