Roping in Uncertainty: Robustness and Regularization in Markov Games