Fix most mypy attr-defined & pyright reportAttributeAccessIssue by explicitly defining attributes on classes

[Avasam]

Summary of changes

Fix most mypy attr-defined & pyright reportAttributeAccessIssue for #2345 by explicitly defining attributes on classes

I have some ideas on how to address the last attr-defined cause by DistributionMetadatalicense_files DistributionMetadata.license_file, but that's out of scope for this PR.

get_finalized_command, reinitialize_command, get_command_obj and get_command_class still require casting, but that should be fixed in typeshed.

Pull Request Checklist

• Changes have tests
• News fragment added in newsfragments/.
  (See documentation for details)

[Avasam]

python/typeshed#12467

[abravalheri]

There is one very edge-y case where it is useful to not define default values beforehand: to avoid people forgetting to call finalize_options if they modify the extension list like in #4529.

Tricky... In this case I feel like would like to have the cake (static type annotations) but eat it too (have the exception happening if someone forgets to call finalize_options).

[abravalheri]

Maybe wrap in a if TYPE_CHECKING and just use the type annotations instead of assigning a value?

[Avasam]

One could also declare the attribute type without a default value:

>>> class A:
...   foo: str
...   bar: str
...
>>> A().foo
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
AttributeError: 'A' object has no attribute 'foo'

If it's an issue with static checkers, then it can be guarded by if TYPE_CHECKING. Either way I'll add a comment that a potential AttributeError is expected and wanted. You can have your cake and eat it too, here 😄

[Avasam]

New missing coverage https://github.com/pypa/setuptools/actions/runs/10306778328/job/28530606282?pr=4535#step:5:108

setuptools/tests/init.py (66.7%): Missing lines 12

Is better resolved by jaraco/skeleton#130

[abravalheri]

Thank you!