fix: Cleanup prometheus metrics for dead workers

Issue: Prometheus gauges not working properly (values from dead processes taken into account)

The current "cleanup" function that is removing orphan Prometheus files is actually never invoked.
Let's move this to a gunicorn config file (which is loaded by default) to register our hook(s).

https://docs.gunicorn.org/en/stable/configure.html
https://docs.gunicorn.org/en/stable/settings.html#child-exit

Signed-off-by: moson moson@archlinux.org