Maneuver Decision-Making Through Automatic Curriculum Reinforcement Learning Without Handcrafted Reward functions