gh-104683: Argument clinic: cleanup state_modulename_name()

[AlexWaygood]

The state_modulename_name() function has two variables named module, that have different purposes and that are assigned to objects of different types:

cpython/Tools/clinic/clinic.py

Line 4718 in 6e850c3

module, cls = self.clinic._module_and_class(fields)

cpython/Tools/clinic/clinic.py

Lines 4746 to 4750 in 6e850c3

	module = None
	try:
	module = ast.parse(ast_input)
	except SyntaxError:
	pass

If we give the second variable a different name, it makes the code easier to understand for mypy and for humans. I also made some incidental changes to nearby lines to use f-strings rather than string concatenation using +.

(This PR is required in order to add type hints to the last remaining untyped function in clinic.py, _module_and_class().)

• Issue: Modernise code in Tools/clinic/ #104683

[erlend-aasland]

This is a very nice improvement, indeed. A lot of the error paths in this PR don't have test coverage, yet. IMO, we should bring test coverage of state_modulename_name() up to 100% before continuing with this PR. I can give that a stab tonight, unless someone beats me to it :)

[erlend-aasland]

Thanks!

[erlend-aasland]

(Sorry, I forgot to mention: we should add tests in separate PRs, since we normally want those backported to bugfix branches.)

[AlexWaygood]

(Sorry, I forgot to mention: we should add tests in separate PRs, since we normally want those backported to bugfix branches.)

In general, I agree -- but here, adding tests revealed some micro-bugs in the failure messages, so I would have had to assert arguably incorrect behaviour in the tests in order to backport them to bugfix branches. Unless we consider the microbugs in the failure messages important enough to warrant backporting this PR as a bugfix, which I personally don't? Anyway, that was my thought process :)