Attempt to fix pydantic-model-prototype tests

[talonchandler]

This PR is my attempt to fix the tests on the pydantic-model-protoype branch.

@ieivanov demonstrated his local failure to me, and pointed out that napari seemed to hang. Although I'm not able to reproduce @ieivanov's hang locally myself, his demo was a key lead---when I skip test_dock_widget (see d41f4c5), the tests pass.

This is a "cheating" fix, though, because this is a useful test. @amitabhverma I now suspect that that your first failing commit f91b4e9 introduced a regression that is interfering with napari's shutdown/cleanup. I tried a couple easy ideas (see notes below), but I wasn't able to find the core issue. Is it possible that the jobs manager isn't cleaning up correctly? I have not read through every line of f91b4e9 yet.

Notes:
d41f4c5 passes --- this is a "cheating" fix, which skips the napari test. This points me towards hanging processes that are interfering with the napari process.
358d514 hangs --- this rules out the cli debugging path as being the issue.
e0e2715 hangs --- this rules out the tab_recon debugging path as being the issue.
9b71e19 hangs --- confirming that the HEAD of pydantic-model-prototype hangs

[amitabhverma]

Thanks @talonchandler
This was very useful in finding the issue which seem to arise from the server socket blocking napari closing when running headless for tests. I believe the close event is not triggered in this case which is where I do the cleanup and stopping of threads/socket/processes currently running.

The changes have been incorporated #487

[talonchandler]

Great! Closing.