Join node deduces the distribution of its output tuple based on
the output tuples of its left and right children. The logic of
deduction was broken: if we swap the children the distribution
changed. So, current commit rewrites join distribution deduction.

Also we refactor some parts of the set_distribution() function.
Its code style is still bad, but now it is not so horrible as
before.

Planner may split some query into range of sub queries, where subquery
execution depends on results previous subquery. Then planner analyzes
query and if it finds similar situation he appends `motion` node to plan.
If executor finds `monion` node in execution plan, he executes
motion sub-query and saves result to virtual table, which linker will be used.

Also fixed the grammar. Previously, we checked keywords incorrectly
and sometimes get false positive "invalid command" errors.

Extract `Engine` interface for working with different execution engine.

Old code was removed. Integration tests was adopted to new executor