implement sql alchemy destination

[rudolfix]

Background
We can handle many rdbmses with a single SQLAlchemy destination. In this ticket we want to implement a generic loader and test it with two engines: SQLLite and MySQL

Requirements

1. implement bulk inserts if possible

• we may need a new file writer from which sql alchemy inserts are the fastest
• some dialect support bulk inserts
  https://stackoverflow.com/questions/3659142/bulk-insert-with-sqlalchemy-orm

2. sql alchemy has identifier length in each dialect
3. now destination capabilities may be dynamically modified. Use the connection string (that contains engine+dialect) to synthesize it.
4. if needed add sql_alchemy_caps to destination caps with additional settings (ie. if bulk inserts are supported or JSON fields)

• allow user to provide own sql_alchemy_caps

5. Use SQLALchemy to create and reflect databases. Implement sql_client interface (but probably form scratch).
6. write dispositions:

• implement append and replace write disposition
• merge in next ticket (I’m not even sure it is possible. But we are using only INSERT and DELETE so maybe?)

Tests:

• most of the tests that abstract sql_client and job_client should pass right away (ie. pipeline tests, state sync, drop command)
• sql_client tests should also pass