Fix database migration edge cases

[piercefreeman]

This PR mainly addresses edge cases with our experimental data migration plugin.

• Add additional unit tests to check expected transformation when executed against Postgres. This validates the accuracy (and syntax) of our templated SQL queries.
• Workaround for dropping enums from existing table columns. Enums in Postgres can't be modified in-place, so the best workaround is to recreate them and then drop the old values. We now keep track of where enum values are referenced so we can migrate these to a new enum object that will replace the old one.
• Add support for "check" constraints.
• Modify a group of enum values at the same time (ie. pass all different enums) to avoid verbose migration files and repeated SQL commands to create/drop old enum types.
• Allow reserved table names within identifier strings (like a "user" table separate from the postgres "USER" command).

We also add some initial docs on getting started with the migration pipeline.