Error if the --python option is specified after the subcommand name

[pfmoore]

Fixes #12067

[uranusjr]

I wonder if we should fail instead of warn since this is likely not going to do what the user wants, and the unintended side efects may be difficult to revert.

[pfmoore]

I also wondered about that. I'm happy to change this.

[pfmoore]

Hmm, test errors look like test_install_nonlocal_compatible_wheel wasn't using the right option name for --python-version 🙁

I've corrected the option name in that test, but it's a bit worrying that the test didn't actually fail previously. If someone wants to fix that, they can do so in a followup PR, though - I think the feature that test was added for probably needs more comprehensive tests, or alternatively needs to try to test fewer things in a single test...

[pfmoore]

OK, this seems good now, moving out of draft status.