refactor: normalize reward range to [0.01, 0.99] and standardize episode scoring key to score across environment and inference logic

feat: enhance safety moderation with keyword-based fallback, robust LLM response parsing, and unified UI insight formatting.

refactor: update submission validator with strict log format checks, HF_TOKEN safety verification, and add FastAPI server for interactive task testing

UI: Move RUN MODERATION button above PURGE LOGS button

Configure app to use port 7860 for Hugging Face compatibility

Clean initial commit without binary files