Maximum Entropy-Regularized Multi-Goal Reinforcement Learning