Add admin handler to list of handlers used for background tasks

[H-Shay]

Fixes #17823

While we're at it, makes a change where the redactions are sent as the admin if the user is not a member of the server (otherwise these fail with a "User must be our own" message).

Dev notes

Keywords:

• TaskScheduler
• run_as_background_process(...)
• wait_for_background_processes(...)
• reactor

[H-Shay]

This is a very janky way of saying "please main process wait until the worker process has finished executing the redaction task before running the rest of the test code." I would love to know the actual way to do this.

[MadLittleMods]

This sort of thing might be good enough. For example, look at how wait_for_background_updates() works.

Perhaps just change this to a while loop with a timeout (example in wait_on_thread()).

Could be updated to use the TaskScheduler functions directly instead of the HTTP requests but keeping it as a black box test is also great 👍

One alternative is that because this particular background task is producing a bunch of redactions which you end up checking anyway, is to just /sync until you see a redaction for all the events. I don't know if we have something for that in Synapse but Complement has MustSyncUntil for example that we could emulate over here.

---

It looks like the TaskScheduler uses run_as_background_process(...) behind the scenes. Here is a previous PR where I wanted to wait for things that ran in a background process:

• Process previously failed backfill events in the background matrix-org/synapse#15585 (comment)
• Process previously failed backfill events in the background matrix-org/synapse#15585 (comment)

Although, I'm unable to get the trick suggested there to work here. Here is what I tried:

        worker = self.make_worker_hs(...)

        [...]

        # Wait for the next run of the scheduler loop
        worker.get_reactor().advance((TaskScheduler.SCHEDULE_INTERVAL_MS / 1000))
        # Ensure `run_as_background_process(...)` has a chance to run (essentially
        # `wait_for_background_processes()`)
        worker.get_reactor().pump((0.1,))
        # Even adding this for good measure, doesn't work
        self.reactor.pump((0.1,))

As an aside (for my own reference), reactor.pump([1]) vs reactor.advance(1) seem to be the same thing. I wonder why both exist.

[H-Shay]

I've opted to use a while loop - hopefully this is sufficient? I went down the rabbit hole of trying to poke at the reactor before resorting to what I originally submitted, so if the while loop is acceptable let's leave it at that :)

[H-Shay]

Thanks for the review @MadLittleMods! I've made the suggested changes - if you're satisfied can you hit the merge button when you get a chance? I no longer have the power to do so :)