Remove dead code from the test system.

purpose: OOS-build tests fail

purpose: OOS-build tests fail

Important changes:
  * lua_libs are now copy from TarantoolServer (if you need, e.g.,
    to copy .lua or .xlog file when you starts server you may
    simply add it to server.lua_libs variable, and in the end
    simply delete it)
  * server now stops on random ports, for your convinience, when you
    `--start-and-exit` test-system - it shows on what port it starts
  * when you want to create a server, that is replica of some other
    server, you must do something like `replica.rpl_master = master`,
    then all will be ok. if hot_standby, then
    `hot_standby.hot_master = master`. when you stop working with it:
    `replica.rpl_master = None`..
  * new option: server.shebang is used for running scripts, no need
    to spoil server.binary now.
  * for config, init_lua and other options now there are decoreators.
    you may use relative paths now.

Other:
  * Some tests were rewritten for new TarantoolServer
  * Add `environment variables` for lua tests, possible to send it to
    server.
  * add rpl_master and hot_master for server directives in lua_tests
    (better idea, it seems, to move all `replication/` tests to python)

Some issues:
  * --valgrind:
    1. `chase_cuOff: no entry for 0x07422`... It garbages output.
    2. If you redeploy server then valgrind log is overwritten.
       Possible solutions:
         a) `valgrind_disable` this tests.
         b) Append logfile with temporary logfiles.
    3. Some memory leaks: need to be fixed.
  * --gdb: some minor code problems, all works, but tests, that
    crashes server won't work (new system of knowing that server
    is up)
  * Need of minor(!!) cleanup

Allow start of tarnatool_box in #! form.

Remove script_dir.

Make all configuration options optional (primary_port
was the last one required).
Make the configuration file itself optional.

If no watcher is started for the event loop, either based on
the config file settings, or in the #! script, do not start the event loop.

Update the test runner to take the new changes into account.

Fix a bug in symbols_load() when no symbols
could be loaded when path to executable was neither
relative nor absolute (but based on PATH).

Fix a bug with ignored return value of slab_arena_create()
in tuple.cc

Because of v2->v3 migration mess different systems have different python
set by default. It is recommended to use versioned binaries.

Signed-off-by: Anatol Pomozov <anatol.pomozov@gmail.com>

Fix session and bad_trigger test
Fix output (remove garbage, fix whe interrupted by exceptions)
Add list of broken tests when tests are run with "./test-run.py --force" flag (make test-force)

1) Moving ./test-run.py --gdb from xterm (or another standart X terminal emulator) to detached Screen session
2) Moving from unnecessary dependency "pexpect" to builtin "subprocess" module.

Don't drop entire vardir on test suite start, it may have
symlinks, configuration files, etc. - useful for debugging.

1) Temporary result files weren't deleted, if test's skipcond failed.
2) Vardir is removed when tests started.
3) Sort tests and test-suites by name
4) Added cleanup to box/lua.test

2. Refactor find_tests code
3. Rewrite and Add comment for chdir(base) (test_run.py:145)
4. Rewrite .gitignore file for new name of unit tests

In the first place Tarantool has only one kind of test - functional testing.
But, when system began to grow, and each component (inside it) is needed to be tested thoroughly.

* 	Added support of UnitTest without creating of *.test file, that contains "run_test("*")"
	(Added UnitTest(Test) in the unittest_server.py, FuncTest(Test) in the tarantool_server.py.
	Now, when we search for tests we tell server to search them, and when we want to execute all
	tests TestSuite runs `execute` method of specific Test subclass)

* 	Fix of bug, that throwing OSError exception, when running by "python test-run.py",
	instead of "./test-run.py"

1) Delete TarantoolServer class, delegate it's behaviour to TarantoolBoxServer and rename it to TarantoolBox
2) Delete 'module' from .ini file (Not needed anymore, cause of Tarantool/Box is now Tarantool and it has only one backend.
3) Refactoring of code respectivly for changes in 1, 2

Sync the WAL at close time.
Fix a race condition between wal_schedule and ev_schedule
this revealed.

A cancelled fiber could get scheduled by wal_schedule.
Do not allow to asynchronously cancel a fiber which
is a part of WAL scheduler queue.
Add a test case.

Test environment:
 - test-run.py script was reformated to PEP8.

	 generalize Server class
	 implement TarantoolServer class
	 implement TarantoolSilverboxServer class
	 implement TarantoolFeederServer class
	 a lot changes in interfaces and bugfixes

Allow disjunctions in SELECT WHERE clause and LIMIT
clause in SELECT, to be able to cover the problem
reported in Bug#729758.

Fix a bug with prepare_gdb arguments (remove an extra
space in -e option).

Add a simple test for Bug#729758.

The following bugs were reported in the process:

Bug#729789 "Zero limit is treated the same as no limit"
Bug#730593 "Bad data if incomplete tuple"
Bug#730613 "SELECT with a disjunction on multiple keys returns
matching tuples twice"

Rename 'silverbox' to 'box', across the board.
Delete the remnants of the old build system
(scripts/rules.mk and scripts/config-def.mk).

Do not touch client/perl for now, since it's maintained
elsewhere.

Return a non-zero exit status if there are test
failures, even if run with --force. This necessary to
correctly identify a test failure in buildbot run.

Rename a test file to avoid shell escapes, align
a comment, modify ./test-run defaults to follow
the new build layout.

Update help messages, add comments, add more checks,
fix minor bugs.

Use CMake to:
- Compile tarantool_silverbox, tarantool_feeder.
- enable ./test-run testing out-of-source
- implement Debug/Release configurations
- add ENABLE_BACKTRACE ENABLE_TRACE and HAVE_BFD switches
- support auto-generated targets