Clear worker thread cache after forking

[A5rocks]

Fixes #2764

While we don't provide any sort of guarantees and I think it's tempting to give up on this:

• Python (3.12+) will already warn on this, so they have that handled
• It's friendlier to do the right thing.

As long as people aren't initializing their own thread caches, this should be fine.

[A5rocks]

Maybe there should be a way to tell Trio "hey I don't want the worker threads anymore, please kill them!", especially in light of the new deprecation warning (ATM the way to do that is to do time.sleep(10)). Maybe Trio should explicitly support this use case and set up a callback for before os.fork() s.t. all the threads get killed.

That's not what this PR does!

Counterpoint: what if a thread is currently doing work (no guarantees even outside of a trio.run because abandon_on_cancel).

[codecov]

Codecov Report

Attention: Patch coverage is 85.71429% with 3 lines in your changes missing coverage. Please review.

Project coverage is 99.98405%. Comparing base (8a7674c) to head (b146087).

Files with missing lines	Patch %	Lines
src/trio/_core/_tests/test_thread_cache.py	87.50000%	1 Missing and 1 partial ⚠️
src/trio/_core/_thread_cache.py	80.00000%	1 Missing ⚠️

Additional details and impacted files

@@                 Coverage Diff                  @@
##                 main       #3200         +/-   ##
====================================================
- Coverage   100.00000%   99.98405%   -0.01596%     
====================================================
  Files             124         124                 
  Lines           18792       18813         +21     
  Branches         1268        1270          +2     
====================================================
+ Hits            18792       18810         +18     
- Misses              0           2          +2     
- Partials            0           1          +1     

Files with missing lines	Coverage Δ
src/trio/_core/_thread_cache.py	98.88889% <80.00000%> (-1.11111%)	⬇️
src/trio/_core/_tests/test_thread_cache.py	98.46154% <87.50000%> (-1.53847%)	⬇️

[jakkdl]

coverage fail appears to be nedbat/coveragepy#310
seems to be workarounds in there, though this is minor enough we could ofc just # pragma: no cover