Fix opportunity search across skills, tags, and AI tracks

[jmgasper]

Summary

• searches challenge names and descriptions case-insensitively
• resolves standardized skill-name matches to persisted challenge skill IDs
• searches authored tags case-insensitively with partial matching
• treats AI as the synthetic exact AI-tag track facet and unions it with persisted track facets before count and pagination
• keeps unknown-only track facets from silently broadening results
• documents bracket-array and track-facet semantics in Swagger

Validation

• pnpm lint
• pnpm build
• isolated PostgreSQL unit coverage: unified name/description/tag/skill search and AI/real-track union, count, and pagination (2 passing)
• focused HTTP E2E route coverage for tag/skill search and pagination headers (1 passing)
• git diff --check

Database test note

The repository migration chain currently fails at the existing 20251122120000_drop_phase_order_idx_allow_null_ai_reviewer migration because isAIReviewer is absent. The focused tests therefore used a new disposable PostgreSQL database populated with prisma db push --force-reset; it was removed after validation.