Move partitioning backfill migration to post-deploy

The partition_table_by_date helper contains several steps:

1. Create a new table with the same schema
2. Install a trigger to sync new data coming in (and updates, deletes)
3. Schedule background migration jobs to backfill the new table

Step (3) should be moved to a post-deploy migration. The reason for this is no-downtime deployments: A regular migration runs before the fleet's code is being reloaded. We would already schedule sidekiq jobs in that migration and only later reload code. Sidekiq jobs would fail because the relevant code might not yet be available to them or not yet available in the expected version.

This is not a blocker for !36298 (merged) as the code for the background migration shipped well in advance (and in the previous release). We should fix this in %13.2 though, so this doesn't become an issue later.