Tue Jan 14 17:04:00 CET 2020

csp indexed scheduler

It's much more direct now.  Compared to the original search, this
should be a lot faster.  I don't think I need to test speed to know
that I don't need the old implementation any more.

This will be a lot faster in the common case where there are a lot of
blocked tasks and each scheduler run only schedules a couple of tasks.

It does need more memory.  It's itching to optimize this but maybe
measure memory usage first, then see what makes most sense in
practical scenarios.

Also needs an allocator for channels, which is usuful in its own

The hot list can still use the pointer inside the task.