corpus(tier1): expand court-documented legal AI citation failures

[yenk]

Problem

Tier 1 currently contains a small set of canonical legal AI citation failures.
The benchmark should expand carefully using only:

• court-documented incidents
• canonically retrievable authorities
• reproducible source URLs
• stable citation metadata

---

Requirements for scoring-eligible additions

Each record should include:

• authoritative source URL
• retrieval timestamp
• canonical citation metadata
• reproducible verification path
• documented judicial or regulatory context

---

Non-goals

Do NOT add:

• social-media anecdotes
• unverifiable claims
• media summaries without underlying authority
• speculative incidents

---

Potential future additions

• sanctions orders
• judicial findings involving fabricated citations
• disciplinary proceedings
• verified filing incidents involving non-existent authorities