CREATE TRIGGER procrastinate_jobs_notify_queue_job_inserted AFTER INSERT OR UPDATE

[medihack]

Unrelated to this PR, but... shouldn't this trigger AFTER INSERT OR UPDATE ?

Currently what happens is that after a job fails and is retried, the worker needs a new job to work on, so it will request one immediately and usually will receive the retry. But if a worker finishes a job while shutting down, and the job fails and is retried, then the trigger won't trigger (because it's an update and not an insert) and we'll need to wait for longer for other workers to pick up the job.

But then, unrelated to this PR. Marking it here to remember opening an issue.

Originally posted by @ewjoachim in #1172 (comment)