Support multi-language YouTube transcript fallback (en, es, pt)

[23r2efewvcs]

Problem

youtube_yt.py:_fetch_transcript_ytdlp() (line 508) hardcodes --sub-lang en — only English auto-captions are attempted. For non-English content, this returns None and the video appears in results without a transcript.

The direct HTTP fallback (_fetch_transcript_direct(), line 470) already supports multi-language: it falls back to the first available caption track when no English track is found. But the primary yt-dlp path does not.

Current behavior:

• yt-dlp path: --sub-lang en → English only → returns None for non-English videos
• Direct HTTP fallback: tries en → en-* → first available track → works for non-English

Proposed Solution

Change --sub-lang en to --sub-lang en,es,pt in _fetch_transcript_ytdlp(). yt-dlp already supports comma-separated language lists — it tries each in order and downloads the first available.

# youtube_yt.py line 508, before:
"--sub-lang", "en",

# after:
"--sub-lang", "en,es,pt",

Optionally make this configurable via env var LAST30DAYS_YT_SUB_LANGS with default en,es,pt, read in env.py.

LLMs understand all three languages natively. A transcript in any of these is better than no transcript.

Estimated impact: +30-50% more transcripts captured, especially for non-English content.

Alternatives Considered

• Use only en,en-orig — catches original language when auto-caption exists, but misses non-English creators entirely
• Download ALL available languages — wastes bandwidth and storage for marginal gain
• Make it configurable without default — users won't configure it, same problem persists
• Rely on direct HTTP fallback only — fallback is less reliable (no yt-dlp retry logic, timeout handling)