feat: add Facebook video extractor

[eagle1maledetto]

Summary

Adds a new extractor for downloading videos from Facebook, supporting all common URL formats:

• Reels: facebook.com/reel/<id>
• Videos: facebook.com/videos/<id>, facebook.com/<user>/videos/<id>
• Watch: facebook.com/watch/?v=<id> (with v= in any query parameter position)
• Posts: facebook.com/posts/<id>, facebook.com/<user>/posts/<id>
• Share links: facebook.com/share/r|v|p/<id> (resolved via HTTP redirect)
• Mobile: m.facebook.com and mbasic.facebook.com variants

How it works

The extractor fetches the Facebook page HTML using auth cookies (required — loaded automatically from private/cookies/facebook.txt in Netscape format, same mechanism already used by the Twitter extractor) and a Chrome-like User-Agent.

Video URLs are extracted from the embedded JSON data in the HTML by matching progressive_url entries with "quality": "HD" or "quality": "SD" metadata. Both HD and SD formats are provided when available, letting the downstream format selector choose the best quality.

Video-ID-anchored extraction

Facebook pages embed data for multiple videos in a single HTML response (feed recommendations, related content, etc.). A naive global regex match on progressive_url would return the first occurrence, which often belongs to a different video than the one requested.

To solve this, the findVideoSection() function scopes the regex search to the correct video's data block:

1. Locates the dash_mpd_debug.mpd?v=<videoID> anchor — unique per video within the videoDeliveryResponseResult JSON structure
2. Bounds the search to the section ending at "id":"<videoID>"
3. Applies HD/SD extraction regex only within this scoped section
4. Falls back to full-body search for single-video pages (e.g., direct reel links)

Watch URL normalization

/watch/?v=XXX pages return HTML with inconsistent/misleading video data when scraped. These URLs are normalized to /reel/XXX before fetching, which yields reliable single-video pages. This is transparent to the user — they submit a watch URL and get the correct video.

Share link handling

ShareExtractor is registered before the main Extractor to match /share/r|v|p/<id> URLs first. It follows the HTTP redirect to the canonical Facebook URL, then returns it for re-matching against the main extractor pattern.

Files

File	Description
internal/extractors/facebook/main.go	ShareExtractor + Extractor definitions, GetMedia(), buildMedia()
internal/extractors/facebook/models.go	VideoData struct (HD/SD URLs, title, dimensions)
internal/extractors/facebook/util.go	GetVideoData(), parseVideoFromBody(), findVideoSection(), URL/Unicode unescape helpers
internal/extractors/main.go	Registration of both extractors in the global Extractors slice

Compliance with project standards

• File structure: follows the standard 3-file extractor pattern (main.go, models.go, util.go) used by all existing extractors
• Cookie handling: uses the existing private/cookies/<id>.txt Netscape format mechanism (same as Twitter)
• Extractor registration: added to internal/extractors/main.go with ShareExtractor before Extractor (order matters for URL matching)
• Commit style: Conventional Commits (feat: ...)
• No new dependencies: uses only bytes, regexp, strings, and existing project packages
• No tests: consistent with the rest of the codebase (no *_test.go files exist in the project)

Configuration

Requires a private/cookies/facebook.txt file with valid Facebook session cookies in Netscape format. The following cookies are needed: datr, sb, c_user, fr, xs.

Additionally, it is recommended to enable TLS fingerprint impersonation (impersonate: true) in the extractor YAML config to avoid detection by Facebook's bot protection.