Downgrade V20 should delete tables in configuration

[Morgan Jones]

the v20 migration in trove/db/sqlalchemy/migrate_repo/versions/020_configurations.py does not delete the appropriate tables on downgrade, and instead adds an exception block to ignore exceptions that MAY be caused by "table already exists" type errors.

This has the following consequences:

- Exceptions in initially creating the tables/columns/etc will be ignored causing the upgrade migration to fail, yet the migration will continue with silent errors
- Exceptions in subsequent upgrades may fail for other errors that generate the same exceptions
- Errors caused by code in future versions that corrupt the data in the tables will not be cleared during the downgrade/upgrade process, effectively leaving the user in an unrecoverable state
- The configuration data left in the tables may have business relationships to data in other tables which are cleared during their downgrade, thus invalidating the retained data

For these reasons, the downgrade should drop the appropriate tables and the exception catching should be removed from the update method.

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.openstack.org/117048

[OpenStack Infra (hudson-openstack)]

Change abandoned by Simon Chang (<email address hidden>) on branch: master
Review: https://review.openstack.org/117048

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.openstack.org/117291

[Simon Chang (changsimon)]

cp16net, amcrn, vgnbkr and I discussed on the IRC and agreed that we won't drop the config tables on downgrade for v20. Code will be added in upgrade to detect table existence, create_table operation will be skipped and a warning will be logged if the tables already exist.

[OpenStack Infra (hudson-openstack)]

Change abandoned by Simon Chang (<email address hidden>) on branch: master
Review: https://review.openstack.org/117291
Reason: Will be included in Change #113582.