feat(scheduled flag changes): implement scheduling of changes

[jurajmajerik]

Changes

Add a Celery task that populates scheduled changes on relevant models every 2 minutes. The current implementation works with Feature Flags; more models might be added in the future.

Notes on the implementation:

• Since we have multiple concurrent Celery workers, we want to enforce concurrency=1 to prevent competing updates. To ensure this, we lock all the relevant rows with SELECT FOR UPDATE NOWAIT.
• We process the rows in chunks to limit the memory usage. For now setting the chunk size somewhat arbitrarily to 10,000. If my back-of-the-envelope is correct, this should be on the order of 10s of megabytes of memory at most.
• The changes specified in the payload of the scheduled_change record is populated on the model instance. If a particular update fails, log the reason for the failure in the failure_reason field of the scheduled_change record. This ensures the change is not re-processed.

How did you test this code?

Integration tests, 👀

[neilkakkar]

looking much better :)