Expose independent grader scores in the reward calculator component models

Fix float rounding output limit to prevent out of bounds exactly 1.0 rewards in logged inference metrics

Update baseline scores

Fix Phase 2 requirements: Add / endpoint for HF health checks and tasks grader definitions

Make environment fully checklist-compliant

Completed

Made some changes

AI completed