[roll] Roll fuchsia [kernel][sched] Rework thread queue events

Rework the contents of thread queue events so that

1) The contents attempt to be packed in the lower bits of the 64 bit
   arguments as much as possible. This increases the proability that
   when the are rendered by either chromium trace viewer, or perfetto,
   that they have not been damaged by the conversion from uint64 ->
   double by the javascript engine which is driving the visualizer.
2) They contain bits which give both the eligibility status and
   current thread type (fair vs. deadline) at the time of the event.

These events are typically disabled at compile time, and are only
consumed by my personal trace analysis scripts right now, so there
should be no concerns about breaking tools which depend on them (there
are no public tools which do).

The old lock contention code in here is currently broken, but I lost
the changes which instrument the OWQs on the kernel side of things
anyway, so I'm not too concerned here.

Cleanup and feature tasks moving forward include:
1) Remove the old contention code.
2) Break the code into multiple files.  This single file thing is
   getting unreadable and hard to maintain.
3) Add analysis tools which look for thread flapping (running and
   blocking for very short periods of time consecutively)
4) Add analysis tools which quantify scheduling latency (the time
   spent in the ready state immediately after having been blocked
   before finally being run on a CPU.
5) Reintroduce hooks into the kernel and add analysis code to help us
   understand how lock contention is affecting threads.

Original-Reviewed-on: https://fuchsia-review.googlesource.com/c/fuchsia/+/696450
Original-Revision: 606995773b1706832c8c9a90ead4d7eff1b1611b
GitOrigin-RevId: ecf5fc24c81e52ea06c428820c014699a56526d7
Change-Id: I4454b4080ec1bdfaab2168ee10a384ab72f81b91