Fast Policy Extragradient Methods for Competitive Games with Entropy Regularization