task: module CLI patching methods

[jharlow-intel]

adds python -m mkl_fft patch install/uninstall/status methods, and also python -m mkl_fft with_patch
Convenience CLI functions that will make it way easier to let users have automagic mkl_fft optimizations in their workflows, such as if they want to run npbench suites but with mkl_fft optimizations.

Tested locally pretty exhaustively, everything worked pretty well. A lot of documentation is claude code and it was way too verbose. I deleted and edited a lot of it, but please take a look at it all and make sure it's correct.

If this gets merged, I'll do the same for mkl_random and mkl_umath

[Copilot]

Pull request overview

Adds a user-facing CLI for enabling/disabling mkl_fft’s NumPy FFT patch, aiming to support both persistent patching (python -m mkl_fft patch ...) and one-shot execution (python -m mkl_fft with_patch ...) alongside updated README guidance.

Changes:

• Introduces python -m mkl_fft patch install|uninstall|status for persistent patch management via a .pth file.
• Introduces python -m mkl_fft with_patch <command> ... intended to run a single command with temporary patching.
• Adds README documentation and unit tests for the persistent patch helper functions.

Reviewed changes

Copilot reviewed 5 out of 5 changed files in this pull request and generated 11 comments.

Show a summary per file

File	Description
README.md	Documents persistent patching, one-shot execution, and programmatic patch APIs.
mkl_fft/main.py	Adds python -m mkl_fft ... dispatcher for patch and with_patch subcommands.
mkl_fft/patch.py	Implements persistent patch install/uninstall/status via writing/removing a .pth file.
mkl_fft/with_patch.py	Implements “one-shot” patch execution by spawning a subprocess with modified environment.
mkl_fft/tests/test_cli.py	Adds tests for install_patch, uninstall_patch, and check_status using a mocked .pth path.

[jharlow-intel]

@antonwolfy @ndgrigorian ready for review, I'll open this for the other 2 mkl packages once this gets merged.

Tried to get thorough CLI testing added to the meta.yaml, but conda seems to do something weird with the PYTHONPATH while running the tests so I had to run a lot of integration tests by hand

[ndgrigorian]

I think we should use argparse here, dpctl is a good example
https://github.com/IntelPython/dpctl/blob/master/dpctl/__main__.py

it's a bit cleaner