feat: expand task suite to 22 challenges and update reward signal mechanics

refactor: enforce strictly positive reward range [0.01, 0.99] and update documentation accordingly

changes to rate limit inference

feat: implement ReAct agent architecture with robust JSON parsing and expanded task suite

Restructure to root server/app.py to satisfy validator

updated inference for hf llm api

added requirements and inference fix.

eas inferenece and added hard, mdeium task