Clean up timeout handling

[andyross]

See discussion in #9620 and #8669

The handle_expired_timeouts() mechanism is fragile. The specific bug was patched, but the code is too large and complicated for what it does (trying to build a list of expired timeouts ahead of time before invoking them instead of just doing a while loop on the queue with simpler constraints).

It also needs some work to unify the tickless handling with the normal timeout stuff, which is needlessly divergent.

[findlayfeng]

Here you need to keep next.
Because the current timeout will destroy the pointer to next if it is re-added, continuing to follow the next loop will have unpredictable results.
Just need to remove remove

[andyross]

Timeouts got reworked in the timer rewrite. Closing my own bug, because obviously I think my new implementation is better than the old one.