Fix some false positive for is_non_empty_f_string

[IDrokin117]

Summary

Closes #21048

Fix some false positive edge cases when determining string truthiness. The following cases are now treated as unknown truthiness:

1. Comparison expressions can return any value, even empty strings.
2. F-strings with format specs can produce either empty or non-empty strings:

bool(f"{'mystring':.0}") # Returns False

Additionally, fixed a bug that caused f"{f""!s}" to be treated as non-empty. Since .iter().all() returns true if the iterator is empty, f-strings with no elements (like nested f"{f""!s}") were incorrectly treated as non-empty.

Test Plan

• Added test cases to S604 since Truthiness has no tests of its own

[github-actions]

ruff-ecosystem results

Linter (stable)

✅ ecosystem check detected no linter changes.

Linter (preview)

✅ ecosystem check detected no linter changes.

Formatter (stable)

✅ ecosystem check detected no format changes.

Formatter (preview)

✅ ecosystem check detected no format changes.

[ntBre]

Thanks! This looks good, I just had a couple of small ideas/suggestions.

[ntBre]

Should this just be any instead? I could certainly be wrong, but it seems like we should check if any part is non-empty rather than all parts are non-empty. That would also give the correct behavior for the empty iterator and match the outer any on line 1401.

I tried this locally, and I think it fixed an existing bug in a SIM222 test too.

[ntBre]

Should we move this down to the other complex expressions below? It seems slightly out of place in this block now that the other variants are all true.

[dscorbett]

To be more precise, this could still be true if all the operators in the Compare are the ones guaranteed to return bools (in, is, is not, and not in).