llm: final-review polish — openai max_tokens parity, errBody rename, rule-sync note, honest key placeholder, README inline list; rebuild standalone

llm: inline spectate into standalone build + README usage (batch + spectate)

redesign: incentive-compatible metric + factorial RxGxE + peer rule-following MCTS + swap + fidelity fixes (Discovery/opponent-invariance/C*)

Upload folder using huggingface_hub

initial commit