Self-Rewarding Rubric-Based Reinforcement Learning for Open-Ended Reasoning