Robust Multi-Agent Reinforcement Learning via Adversarial Regularization: Theoretical Foundation and Stable Algorithms Alexander Bukharin Y an Li Yue Y u