Close executor in event loop if interpreter is closing

[mrocklin]

Fixes #6255

Previously this would cause a RuntimeError during shutdown.

test incoming

[mrocklin]

Added test. I think that we used to have such a test. Maybe it was removed? I'd encourage folks to take a look in case it seems bad in some way.

[github-actions]

Unit Test Results

       16 files  ±  0         16 suites  ±0   7h 59m 2s ⏱️ + 4m 12s
  2 745 tests +  2    2 662 ✔️ +  2       80 💤 ±0  3 ❌ ±0 
21 861 runs  +16  20 810 ✔️ +13  1 044 💤 ±0  7 ❌ +3 

For more details on these failures, see this check.

Results for commit 40a21a6. ± Comparison against base commit 223c815.

[fjetter]

I have a slight preference to be explicit here but I'm fine either way. The following works as well, at least for the test

Suggested change

	try:
	await to_thread(_close)
	except RuntimeError: # Are we shutting down the process?
	_close() # Just run it directly
	from distributed.utils import is_python_shutting_down
	if is_python_shutting_down():
	# Scheduling new futures during shutdown is not allowed
	_close()
	else:
	await to_thread(_close)

[mrocklin]

My preference is to ask for forgiveness rather than permission here. I don't have enough confidence about Python shutting down to know all of the stages involved. Honestly I'd also be ok just except Exception here. The worst thing that can happen is that we call close again. I'd rather take the sledgehammer than the scalpel approach here.