Refactor grader to use openenv.core.rubrics.WeightedSum + Rubric subclasses

Algorithm Distillation: grader v2 with belief_accuracy + SFT pipeline

env: meta-RL refactor (continuous profiles, action+belief, adaptation grader)

Rebuild as Life Simulator: 5 meters, 3 hidden profiles, GRPO training pipeline

Fix bugs, add tests, and improve code quality