Summary

This PR adds subtitle (transcript) extraction support for Bilibili video pages, following the same pattern as the existing YoutubeExtractor.

What it does

When a user visits a Bilibili video page (e.g. bilibili.com/video/BV...), the extractor now:

• Fetches video metadata (title, author, description, publish date, cover image) via the Bilibili view API
• Discovers available subtitle tracks via the player/v2 and player/wbi/v2 APIs (tries multiple endpoints for robustness)
• Selects the best subtitle track based on language preference and track quality (prefers human-authored over AI-generated, respects zh-cn > zh > en fallback order)
• Downloads and parses the subtitle JSON into a structured transcript, with intelligent line grouping (merges short adjacent lines, splits on long pauses) and CJK-aware text concatenation
• Exposes the transcript as a transcript variable (alongside title, author, site, image, published, description, part, language) for downstream consumers like Obsidian Web Clipper

Screenshot

Transcript successfully extracted from a Bilibili video page, visible in the clipper's page variables:

Files changed

• src/extractors/bilibili.ts — New BilibiliExtractor class with async transcript extraction, subtitle track discovery/selection, URL validation (restricted to .hdslb.com / .bilibili.com hosts), and an in-memory LRU transcript cache
• src/extractor-registry.ts — Register BilibiliExtractor for bilibili.com URL patterns

Design notes

• Follows the existing extractor conventions: extends BaseExtractor, implements canExtract / extractAsync / prefersAsync
• Reuses the shared buildTranscript utility for consistent HTML/text output
• Subtitle content fetch omits credentials: include because the Bilibili subtitle CDN (aisubtitle.hdslb.com) returns Access-Control-Allow-Origin: *, which is incompatible with credentialed requests; the URL already carries an auth_key parameter for authorization
• Supports multi-page videos (respects the ?p= URL parameter)
• Protocol-relative subtitle URLs (//i0.hdslb.com/...) are normalized to https://