fix(concat): preserve narration timing and document follow-up

[leether]

Summary

• Rework concat effect-path audio handling so video transitions and narration timing stay aligned.
• Preserve source background audio when mixing TTS, with a low-volume background mix instead of replacing all original audio.
• Add a narration verification helper and document the v1.3 audio lessons plus a follow-up Task Card for pipeline consistency.

Privacy notes

• Removed local absolute paths from the new Task Card before commit.
• Kept the one-off local regenerate_video.py script out of git and excluded locally.
• Did not include generated media, local output directories, tokens, or private URLs.

Validation

• git diff --cached --check
• .venv/bin/python -m py_compile core/concat_engine.py scripts/verify_narration.py harness/self_report.py harness/memory_loader.py harness/harness.py core/segment_tts.py core/timeline_mapper.py core/frame_extractor.py core/cta_resource.py extensions/storyboard/storyboard_ai.py extensions/animations/animation_templates.py extensions/animation_templates/base.py extensions/prompt_templates/base.py
• Staged diff privacy scan for absolute paths, local output names, tokens, and private URL patterns.