feat: Add --whisper-prompt option for custom vocabulary hints

[MarianoFacundoArch]

Summary

Adds the ability to pass vocabulary hints to Whisper's initial_prompt parameter for improved transcription accuracy on brand names, technical terms, and other uncommon words.

Problem

When transcribing audio that contains brand names or technical terms, Whisper often misinterprets them. For example:

• "PlateLens" gets transcribed as "Platelins"
• Brand names and proper nouns are frequently incorrect

Solution

Whisper already supports an initial_prompt parameter that can hint the model about expected vocabulary. This PR exposes that parameter through a new --whisper-prompt CLI option.

Usage

pycaps render --input video.mp4 --template hype --whisper-prompt "PlateLens, BrandName"

Changes

File	Change
whisper_audio_transcriber.py	Accept initial_prompt parameter in __init__ and pass to Whisper's transcribe
caps_pipeline_builder.py	Add initial_prompt parameter to with_whisper_config()
render_cli.py	Add --whisper-prompt CLI option in Whisper panel

Benefits

• ✅ Minimal code change (< 10 lines total)
• ✅ Backward compatible (optional parameter, defaults to None)
• ✅ Leverages existing Whisper capability
• ✅ Significantly improves accuracy for brand names and technical terms
• ✅ No external dependencies

References

• Whisper initial_prompt documentation
• Improving Whisper accuracy with prompts

[francozanardi]

Hey @MarianoFacundoArch

Thank you for this work!! I'll review it and let you know :)