tweak: Drop external foreign key constraints during GTFS import

[jzimbel-mbta]

Summary of changes

Asana Ticket: No ticket, followup to GTFS import feature

In order for TRUNCATE all_the_gtfs_tables to be possible, the tables need to have no dependent foreign key constraints originating from outside the set of tables being truncated.

This achieves that by temporarily dropping such constraints, and then re-adding them after the GTFS tables have been repopulated.

Implementation notes

I wanted to make sure any Arrow -> GTFS foreign key constraints added in the future would be handled automatically by this logic.

I felt it would be too cumbersome for developers to have to remember to update code in two different places only when adding foreign keys satisfying these specific conditions.

As a result the code relies on some introspection tools provided by Postgres and is a bit heavy on custom queries, but I think it's still relatively straightforward, and worth it for the ease of use it provides.

Reviewer Checklist

• Meets ticket's acceptance criteria
• Any new or changed functions have typespecs
• Tests were added for any new functionality (don't just rely on Codecov)
• This branch was deployed to the staging environment and is currently running with no unexpected increase in warnings, and no errors or crashes.

[Whoops]

I don't have a problem with the implementation as written, but how hard would it be to write tests for this functionality? If it's not to much effort, I think it'd be worthwhile to have.

[jzimbel-mbta]

I don't have a problem with the implementation as written, but how hard would it be to write tests for this functionality? If it's not to much effort, I think it'd be worthwhile to have.

@Whoops Sounds good, would testing the functions in ForeignKeyConstraint be enough?
Or would you also like to see tests on Arrow.Gtfs?

The latter is a bit more challenging in its current state, since it's hardcoded to act on the GTFS tables specifically, but I should be able to tweak it to be more testable.

[Whoops]

I think ForeignKeyConstraint is sufficient. If you did try to cover GTFS I'd suggest doing integration tests that import a feed and check on the state of the world afterward, but that may be more work than it's worth.