fix: address CodeRabbit feedback and P0 blockers

feat: implement HFInferenceJudgeHandler for free-tier AI analysis

fix: wire EmbeddingService to simple orchestrator + improve search quality

feat: add demo mode warnings and BYOK (Bring Your Own Key) support

fix: pass API keys explicitly to pydantic-ai providers

fix: pydantic-ai v1 compatibility and Anthropic model update

fix: add type ignore comment for Agent instantiation in JudgeHandler

fix(tests): resolve mypy divergence and test failures

feat(phase3): implement judge slice (LLM Judge, Prompts, Models)

feat(phase1): Foundation & Tooling complete

adds scaffolding