gh-94909: fix joining of absolute and relative Windows paths in pathlib

[barneygale]

Have pathlib use os.path.join() to join arguments to the PurePath initialiser, which fixes a minor bug when handling relative paths with drives.

Previously:

>>> from pathlib import PureWindowsPath
>>> a = 'C:/a/b'
>>> b = 'C:x/y'
>>> PureWindowsPath(a, b)
PureWindowsPath('C:x/y')

Now:

>>> PureWindowsPath(a, b)
PureWindowsPath('C:/a/b/x/y')

• Issue: pathlib: inconsistent handling of rootless Windows paths #94909

Automerge-Triggered-By: GH:brettcannon

[barneygale]

Hey @serhiy-storchaka, would you be willing to review? You fixed the same issue in joinpath() (#64107) almost a decade ago! Thanks.

[brettcannon]

@barneygale if @serhiy-storchaka doesn't review in a week I will go ahead and merge; feel free to ping me to do it if I forget (I'm also stalling because 3.11.0.rc1 is being cut today and I don't want to have to remember to merge into 3.11 once the branch opens again and then forget). 😅

[serhiy-storchaka]

Wait a moment. This change goes against the initial approach. It can be slower and inconsistent with other pathlib code (the definition of "drive" was different in pathlib and os.path, and can still be different).

[barneygale]

Could you give an example where its inconsistent with other parts of pathlib?

It might be slower (I haven't checked) but I don't think pathlib, as a high level path library, should care about beating the performance of os.path for equivalent operations. Comparable performance is fine.

[barneygale]

We could de-risk this by skipping backporting. It's a niche bug, and I only need it solved in main.

[brettcannon]

To keep things simple I only did this to main. We can always backport later if we find it necessary.