feat(tts): add segment-level narration audition lab

[RayJiang4S]

Summary

• add tts_segment_lab, a segment-level narration audition tool that runs before final TTS asset generation
• support dry runs, provider-routed generation through tts_selector, reference audio comparison, and selection.json output
• document the workflow in a new core skill and wire it into the explainer script/asset director guidance

Why

OpenMontage already has tts_selector and sample-preview guidance, but there was no structured way to compare multiple voice/parameter/provider candidates for sensitive narration lines or keep the approved choice available for downstream asset generation. This tool makes that process explicit and repeatable.

Testing

• /Users/ray/Repositories/persional/video-ai-lab/tools/OpenMontage/.venv/bin/python -m pytest tests/tools/test_tts_segment_lab.py tests/contracts/test_phase3_contracts.py -q

[RayJiang4S]

Added a screenshot of the new compare.html review UI.

The latest update also makes the review loop iterative rather than assuming a fixed second round: annotate now reports review_complete and next_operation, apply_review can be repeated until every segment has an approved take, and selection.json is only written after the review is complete.

It also safely auto-loads local .env files for provider credentials and records only the loaded file paths, never secret values.

[RayJiang4S]

Update from the v4 production retrospective:

• Surface failed generated candidates directly in compare.html with provider error text and attempt count, so missing audio is not mistaken for a usable provider sample.
• Add merge_retry operation to merge successful retry/follow-up results back into the original results.json, review.md, and compare.html.
• Document the retry merge workflow in skills/core/tts-segment-lab.md.

Validation:

• .venv/bin/python -m pytest tests/tools/test_tts_segment_lab.py -q -> 21 passed
• .venv/bin/python -m pytest tests/tools/test_tts_segment_lab.py tests/contracts/test_phase3_contracts.py -q -> 84 passed
• .venv/bin/python -m py_compile tools/audio/tts_segment_lab.py tests/tools/test_tts_segment_lab.py