fix: increase grader score epsilon to survive .3f stdout formatting

fix: clamp grader scores to strict (0, 1) interval for Phase 2 compliance

fix: add activity gate to graders, expand README, improve inference

fix: improve env robustness, grader fairness, and inference compatibility

feat: complete Phase 3 — inference, expert task, deployment

feat: implement Phase 2 — environment, graders, server endpoints, Dockerfile

fix: close Phase 0/1 gaps — conftest.py, critical path computation

feat: implement Phase 1 core engine — models, DAG, agents, rewards, tasks

chore: scaffold workflow_orchestrator environment via openenv init