feat: convert to MCP tool-calling pattern matching reference repos

feat: major upgrade — multi-step MDP, AST injection, 5-signal reward

KW-WM proof: reward prediction beats baselines by 38.8-71.7%

Fix all 4 remaining issues for hackathon

Update baselines with real post-fix LLM scores

Fix 4 critical grader bugs

Update baseline with real GPT-4o-mini scores

Improve grader quality + research framing

CodeReviewEnv v1.0 — OpenEnv-compliant submission