This patch introduces two tests:

1. `couple.test_failover` reproduces heatbeat timeout on a follower
   that leads to a new election.

2. `threesome.test_leader_dispuption` simulates disconnected follower.
   It shouldn't disrupt the leader in this case.

Leader death is already tested in `threesome.test_log_rollback`.

Close https://gitlab.com/picodata/picodata/picodata/-/issues/25

Also, fix the code to pass the test:

1. Use `replace` instead of `insert` when appending entries. They may be
   overridden.
2. Increase pool worker queue size. Raft node may send several messages
   within a single tick. Without this change the second message is
   dropped.

The second change is a poorly designed workaround. Constant limit
doesn't seem to be both cost-effective and reliable at the same time,
but we don't have a better solution at the moment.

Close https://gitlab.com/picodata/picodata/picodata/-/issues/26

It makes testing easier. Monkey patching the handler isn't necessary
anymore. Instead, it's mocked now.

Follower makes a proposal. It's applied on both instances.

This patch also introduces a new API `picolib.raft_status`.

Close https://gitlab.com/picodata/picodata/picodata/-/issues/24

+ tarantool-sys submodule
+ tarantool-patches directory
+ build.rs build script to patch and build tarantool
+ static linking with tarantool
+ refactoring for cli arguments

Close https://gitlab.com/picodata/picodata/picodata/-/issues/23

This patch allows two nodes communicate over `net.box`. It implements C
stored procedure `raft_interact` that is called on the server side.

Close https://gitlab.com/picodata/picodata/picodata/-/issues/17

This patch introduces `traft::ConnectionPool` for async nodes
communication.

Close https://gitlab.com/picodata/picodata/picodata/-/issues/18

Now it's possible to set up initial topology via command-line arguments.
Start one of several nodes (for example, two):

```bash
picodata run --peer 127.0.0.1:3301 --peer 127.0.0.1:3302
```

Single-instance configuration is started as before:

```bash
picodata run
```

Module `traft::storage` doesn't panic anymore. Instead, it gracefully
returns errors. All unwraps are moved to the calling functions.

Also, the code is supplemented by tests, and debug logs are removed.

1. Merge `on_ready` handler with the `loop_fn` body into
   the larger `raft_main` function.

2. Get rid of debug logs.

1. Prefer using `"str".into()` since it's briefer.
2. Preserve `"str".to_owned()` if type inference can't help.

1. Fix PICOLIB_AUTORUN env var naming.
2. Add an assertion that it works as expected.
3. Show output if listing tests fails.
3. Run inner tests with listen port :0 to avoid
   "address already in use" errors.

It's a part of networking development.

Also, this patch refactors error handling introduced in previous commit
(see 96aba97a). A single struct is replaced with the enum, common for
all `try_from` functions.

1. Replace `From` trait for raft::Entry with the `TryFrom`. Fairly
   handle its errors in the storage.

2. Replace some other unwraps with "expect" with a more clear message.

3. Support all of `raft::EntryType` variants.

Run it:

```
PICODATA_RAFT_ID=2 cargo run run
```

This is a pre-requisite for further two-node configuration development.
This is a temporary hack until fair joining is implemented.

Close https://gitlab.com/picodata/picodata/picodata/-/issues/16

1. Trait `From` for `traft::row::Entry` now accepts a value instead of a
   reference in order to avoid unnecessary clones. A caller makes a
   clone itself if necessary.

2. Trait `From` implementation for `traft::Message` now accepts a
   reference instead of a value. Anyway, it uses msgpack under the hood
   and doesn't affect performance.

1. Remove `get_stash` function - it doesn't show anything interesting.
2. Adjust logging levels.

Also split log level initialization away from schema.

For ambiguous concepts (such as Entry, Message) use prefix notation instead of suffixes.

This patch eliminates raw raft node dereferencing and redundant use of
`borrow_mut()`.  Instead, `traft::Node` provides a safe API that
communicates with the raw node through the fiber channel.

Also, functions `node::new` and `node::start` are merged because keeping
them distinct makes no sense.