Fix traceback in threads when SetTrace used.

[mickp]

Testing

The following script returns normamly, where previously it encountered: ValueError: call stack is not deep enough.

import sys
import threading
import typing

def tracer(frame, event, arg):
    return tracer

def do_check():
    isinstance(None, typing.Iterable)

sys.settrace(tracer)

t = threading.Thread(target=do_check)
t.start()
t.join()

[mickp]

Resolves #2072

[mickp]

I think the CI tests are failing here, for each configuration.

I see that fail intermittently when I test locally.

When I attach the debugger, the pass rate may be higher, and I see occasional failures at the location above, and here:

[slozier]

Thanks for the PR! Will be a couple weeks before I have time to properly look at this.

[mickp]

I think this addresses the root cause of the call stack loss in threads, but I can not figure out how the change is getting hit to cause that one test failure in all runtime targets.

[slozier]

@mickp I finally got around to looking at this, apologies that it took a while. It was an issue with counting listeners. I pushed a fix along with some tests.

[slozier]

Thanks for the PR!