Fix custom joke sound playback

[ded-furby]

Closes #2

Summary

• play the selected joke sound asset instead of always triggering the generic alarm
• fall back to the existing alarm when file playback is unavailable
• add a regression test for filename-based playback and pin the existing alarm test to the Linux branch it validates

Testing

• python -m pytest -q

[AkshajSinghal]

Hey! Really appreciate the effort! The logic in play_audio_file looks solid, and the fallback to play_alarm() ensures we don’t break existing functionality.

In pyfunny.py, if play_audio_file succeeds but somehow raises an exception after starting playback, we might hear both the custom sound and the default alarm. To be safe, maybe we can add a simple flag or move the play_alarm() call strictly into the except block?

[AkshajSinghal]

Thanks for the fix. Please address these three points before merge:

1. Prevent double playback by ensuring play_alarm() is strictly inside the except block.
2. Explicitly convert paths to strings (str(sound_path)) in play_audio_file to ensure compatibility with all system subprocess calls.
3. Add a note to the README that .mp3 and .wav are recommended for cross-platform support across the different audio modules.

[ded-furby]

I traced the current CI failure to the workflow setup rather than the custom-sound changes themselves. The Python CI job installs the package with python -m pip install -e ., then immediately runs python -m pytest, so the job fails before exercising the branch because pytest is only in the dev extra.

I verified locally that this branch passes with:

python -m pip install -e ".[dev]"
python -m pytest -q

That ran cleanly here: 21 passed in 0.36s. Updating the workflow install step to include the dev extra should unblock the PR checks.