feat(bots): expand AI_BOT_PATTERN to cover 26 more crawlers

[Gdewilde]

Summary

• Expand `AI_BOT_PATTERN` from 23 → 42 tokens, covering Claude-SearchBot/Claude-Web, plain Applebot, Google-CloudVertexBot / Google-Agent / GoogleAgent-Mariner / Gemini-Deep-Research, Amzn-SearchBot / NovaAct, AzureAI-SearchBot, meta-externalfetcher / meta-webindexer, DeepSeek, PanguBot, Webzio-Extended / omgili(bot), Timpibot, all Grok / xAI variants, Manus-User, quillbot, MyCentralAIScraperBot, cohere-training-data-crawler, and Ai2Bot-Dolma.
• Fix the plain `Applebot` gap (regex now uses `Applebot`, still substring-matches `Applebot-Extended`).
• Add `parseBotName` labels for the new vendors (DeepSeek, Huawei, Webz.io, Timpi, xAI, Manus, QuillBot, Microsoft, MyCentralAI, plus new Claude / Google / Amazon / Meta variants).
• `Claude-Code` is intentionally still excluded from `AI_BOT_PATTERN` (per existing doc comment) — it's a coding-agent UA handled by the HTTP-library heuristic.

Brings coverage to 47/48 of the bots tracked by Peec AI's reference list.

Test plan

• `npx vitest run` — 161/161 passing (added 27 new `isAiBot` cases + 25 `parseBotName` cases).
• `npx tsc --noEmit` clean.
• Spot-check a real-world UA log after release to confirm no regressions in browser/HTTP-client classification.

🤖 Generated with Claude Code