feat: environment redesign — real CSV data, shaped rewards, difficulty tiers

fix: actionable hints now include real IDs (location_ids, product_ids, rules)

fix: fallback to max_completion_tokens for newer models

fix: coerce string prices to float in update_variant

fix: inference.py runs all 3 tasks with proper [START]/[STEP]/[END] per task

feat: add /tasks and /grade endpoints for competition validator

fix: move all files to repo root (HF Spaces requires README.md at root)

Fix inference.py stdout format to match sample spec exactly

Add 3 tasks with graders and clamp scores to (0, 1)

Align env var names with OpenEnv sample inference spec

Initial commit: Shopify store audit OpenEnv environment