running prune thread directly in gunicorn

[missytake]

I figured out a way to run threads from the gunicorn main process. This is a prerequisite for #36. We can use this to run all mailadm threads in the same docker container.

Reference:

• Running gunicorn with threading benoitc/gunicorn#2320 (comment) recommends using gunicorn hooks
• I decided that this one fits better than post_worker_init: https://docs.gunicorn.org/en/stable/settings.html?highlight=post_worker_init#on-starting

I added a small watcher thread which will kill the whole process/container if the prune thread dies. This can be extended to the bot thread when we add it in #23. Also, all our threads are daemon threads - together, this ensures that the whole container dies if any of the gunicorn, prune, or bot threads fails, and can be restarted by docker.

In the long term I want to move the app.prune() function to commands.py or something, but this doesn't exist yet on master, so I just added it to app.py for now. It's a bit against DRY, but let's do this properly after #23 is merged.

[missytake]

I tested it with docker already:

• the prune thread starts
• it also exits on error (I moved the DB away to kill the prune thread, that killed the container as well)
• except with a keyboard interrupt I couldn't check whether everything dies when gunicorn dies so far.
• who watches the watcher?

[missytake]

Closing this in favor of #45, actually this is much more useful in the bot development than on master right now.