Resolves #10283; check for module level ignore_missing_imports on missing stubs

[TheCleric]

Description

Fixes #10283

For missing type stubs for legacy bundled packages, we will now look for module level overrides for ignore_missing_imports. If present, we will not issue an error like: Cannot find implementation or library stub for module named "redis.sentinel"

Test Plan

3 new test scenarios were added for this PR in check-modules.test for a missing type stub for a legacy bundled package:

Test Name	ignore_missing_imports global	ignore_missing_imports module override	Result
testIgnoreMissingImportsGlobalWithMissingStub	True	False	Error
testIgnoreMissingImportsWildcardWithMissingStub	False	True	No Error
testIgnoreMissingImportsGlobalAndWildcardWithMissingStub	True	True	No Error

The first two tests were passing before the code change and are there to help ensure no regression. The last test was the use case from #10283 to make sure the new code accomplished its mission.

[JukkaL]

This seems potentially dangerous, since Options can have significant state (_per_module_cache) and it's hard to reason about the performance and correctness implications of the cache. I came up with a simple alternative approach that doesn't use deepcopy, and since there is not much time before the release, I'll create a PR now (but I'll give you credit for the first implementation).

[JukkaL]

Thanks for the PR! I'm closing this in favor of #10582 (see my comment above) primarily since options cloning has previously been a real performance bottleneck for very large projects, and I want to be careful not to regress performance.

[TheCleric]

Thanks for the PR! I'm closing this in favor of #10582 (see my comment above) primarily since options cloning has previously been a real performance bottleneck for very large projects, and I want to be careful not to regress performance.

Completely understand. I wasn't super happy with my approach anyway. 😂