Average-Reward Reinforcement Learning with Entropy Regularization