feat: Add transcript save/reuse with automatic detection

[stratus-ss]

This PR adds the ability to save and reuse speech recognition transcripts, dramatically improving workflow efficiency when processing the same audio multiple times.

Features Added

Core Functionality:

• --save-transcript: Automatically saves generated transcripts as JSON
• --input-transcript: Explicitly loads a saved transcript
• --force-retranscribe: Forces new transcription even if cached transcript exists
• Automatic detection: When --save-transcript is used, existing transcripts are automatically detected and reused

As per the readme you can test with both the included test or

# First run: generates transcript (slow)
monkeyplug -i input/Witch_mother1.m4b -o output/clean.m4a --save-transcript

# Second run: automatically reuses output/clean_transcript.json (fast!)
monkeyplug -i input/Witch_mother1.m4b -o output/clean.m4a --save-transcript

# Use different swear list with same transcript
monkeyplug -i input/Witch_mother1.m4b -o output/strict.m4a \
  --input-transcript output/clean_transcript.json -w swear_list.json

[stratus-ss]

This PR needs to be merged before I can proceed with other PRs so I will hold here in case there are adjustments that need to be made

[mmguero]

I'd prefer to not have binary files in the repo. If you'd remove this file from the PR (I know your test uses it; so if you'd like to keep it in your fork or something that's of course your choice) I'll go ahead and pull it.