More predictable timeout-notify with slow sync calls

[ollanta]

Summary

The timeout-notify mechanism is less predictable than necessary when slow synchronous calls are run on the event loop thread. This can lead to uvicorn workers being killed by (for example) gunicorn while they're still responsive.

The uvicorn servers main_loop triggers on_tick and sleeps asynchronously for 0.1 seconds. Every 10 ticks on_tick checks if it's time to run cfg.callback_notify. The comments in the code imply that the check is supposed to run once every second, but if there are synchronous calls blocking the asynchronous sleeps from triggering in a timely manner, the actual cadence can be a lot slower.

Of course, you're not supposed to block the event loop like that, but checking the need to notify on each tick (every 0.1+ seconds) improves the situation at a negligible cost.

Checklist

• I understand that this PR may be closed in case there was no previous discussion. (This doesn't apply to typos!)
• I've added a test for each change that was introduced, and I tried as much as possible to make a single atomic change.
• I've updated the documentation accordingly.