Introduce SQL statement abstraction framework

[jamadden]

This accomplishes many things:

• Makes it simple and easier to prepare statements, if desired. This replaces a whole bunch of nastiness spread through the MySQL and PostgreSQL code bases. That was the initial reason for this: a number of the polling statements needed to be prepared for improved performance (Rethink invalidation polling #159), but it was getting really ugly to deal with all the combinations. This made it easier to do so.
• Abstracts parameter handling across drivers. This can eliminate many of the "meta variables" we (confusingly) use in conjunction with ScriptRunner
• By having a schema definition in a single place, we can ensure all databases are consistent without editing three copies of schema.py
• By having the schema definition represented in objects, we get import-time query validation when schemas change.
• Simplifies dealing with the differences between history-free and history-preserving databases.
• Help enforce certain query practices (e.g., Make the query abstraction layer enforce all 'FOR UPDATE' queries include 'ORDER BY' #282 )
• Move a lot of datatype conversions (like driver.state_to_bytes() and the like) to a lower level to stop cluttering the main code (not implemented yet).

Ultimately this will mostly replace *_query_property and the use of ScriptRunner and lots of schema.py. That's only partly done here, though.