Independent Policy Gradient Methods for Competitive Reinforcement Learning Constantinos Daskalakis Dylan J. Foster Noah Golowich