Improve hooks tests

[EliahKagan]

This improves the tests of git hooks in test_index.py:

• f664a0b fixes and adds missing xfail marks for the WinBashStatus.Absent condition. This is a test bug from when I added the xfail marks in #1745. The Absent condition was detected correctly, but not always used, or used properly, in xfail markings.
• e148647 adds a direct test of run_commit_hook itself. The other tests test specific kinds of hooks. run_commit_hook is public so it's reasonable for it to have its own test, plus I anticipate this will help future work.

Their commit messages include some more details, including rationale.

I have made sure to test on a Windows system that truly does not have bash.exe (as well testing the WinBashStatus.Wsl case, which CI confirms, and the WinBashStatus.Native case).

One might worry that I missed this before due to some further bug, but I believe not. Producing a genuine WslBashStatus.Absent condition on a Windows system that has a system-provided WSL-related bash.exe in System32 (which is searched regardless of what is in the PATH environment variable, as noted in the WinBashStatus.check docstring) is a bit tricky because renaming that file requires one to act as the TrustedInstaller user (as would the worse option of deleting it). At some point while working on #1745, I did not test this when I should have. I believe that is the only reason it was missed, rather than any underlying technical issue.