Two-decimal scores, single tasks/graders.py, min three inference episodes

Validator delegates to validate-submission; three grader entry files for six tasks

refactor: single tasks/graders.py for all tasks, remove thin grader modules, add .env.example

feat: six canonical tasks, task_registry, DaddyCoder-style graders layout

feat: per-task grader entry files, distinct paths for validators, assets/outputs layout

feat: canonical tasks/graders.py layout for Phase 2; wire inference and manifests

fix: per-task grader modules, tasks manifest, verify script, unittest

Deploy OpenEnv to Space