Consider making spline act as its own deployment nexus

[eevee]

trafficstars

I keep referring to WSGI things running independently as "app processes", but they're hilariously limited: they only sit around in a wait loop until a particular kind of event comes in, then respond to it. There are all manner of hacks around for doing more work "after" sending the response, as though this were a difficult feat to accomplish, and the Real Way is to write a totally separate stack that listens to a message queue and does work.

It's all the same code, and an app process has control over itself, so perhaps spline should support launching both WSGI workers and general-purpose workers.

How exactly this is accomplished is not obvious, hence this ticket. Spline should absolutely not contain an entire WSGI↔HTTP stack, but it might be able to use facilities of whatever's running it: at the very least, both gunicorn and uWSGI have in-Python hooks available. Could write adapters for both.

Or maybe this is a thing mamayo should be responsible for?