Improve media queue chain for video ASR

[jian45154]

Summary

• Mark imported videos as not_processed so Xiaohongshu/WeChat video rows enter cache, queue, and ASR handoff flows.
• Add a Vibe ASR preset alongside Sona and document Vibe usage.
• Normalize bare forwarded social URLs like xiaohongshu.com/..., xhslink.com/..., and mp.weixin.qq.com/....
• Prevent video-only presets from being applied to image OCR commands in generated auto-queue-next handoffs.

Verification

• pytest -q → 238 passed.
• python -m build → wheel and sdist built.
• Hermes local wrapper E2E with Xiaohongshu sample → ingest OK, media cache OK, media-pipeline OK, export verify OK, auto-queue-next verify OK.

Notes

• OCR/ASR execution remains review-gated: this PR prepares verified auto-queue-next.commands.txt; actual Vibe/Sona execution still requires local executable + model.