feat: add external transcript input support to CLI (Issue #11)

[drphero]

Summary

Implements Issue #11 by adding a transcript ingestion path that accepts external transcription files and skips built-in STT, while preserving existing --subtitle-data behavior.

What Changed

• Added new CLI options:
• --transcript <path>
• --transcript-format auto|whisper_json|pycaps_json|srt|vtt (default auto)
• Added CLI validation:
• --transcript and --subtitle-data are mutually exclusive
• --transcript-format requires --transcript
• Added Python builder APIs:
• with_transcription(transcription, format="auto")
• with_transcription_file(path, format="auto")
• Added new transcript parsing layer:
• TranscriptFormat enum
• load_transcription(...) loader
• Supports whisper_json, pycaps_json, srt, vtt
• Includes VTT inline timestamp tag parsing (<HH:MM:SS.mmm>)
• Normalizes all inputs to a pycaps Document with valid word timings
• Updated pipeline execution precedence:

1. subtitle_data path (existing behavior, skips transcription + processing)
2. external transcription (new behavior, skips STT but still processes)
3. built-in transcription flow (existing default)

• Updated docs for CLI, examples, and README.
• Added unit tests for parser coverage, CLI flag validation, and pipeline branching.
• Updated CI workflow to run unit tests in addition to smoke test.

Backward Compatibility

• --subtitle-data semantics are unchanged.
• No new --skip-transcribe flag was added; --transcript implies it.

Test Results

Executed:

.venv/bin/python -m unittest discover -s tests -p "test_*.py"

Result:

• Ran 14 tests
• OK

[francozanardi]

Hey @drphero, thanks for this job!! :) You even included some tests for pycaps.