Update README for unified correct/incorrect schema

Unify schema: each test row has correct_answer/incorrect_answer precomputed; demos use {prompt,answer}

Add explicit configs YAML for dataset viewer

Add flipped_answer JSONL files (one row per item)

Drop old flipped_answer JSON pools (now JSONL)

Upload main eval suite data

Upload README.md with huggingface_hub

initial commit