Quantum reinforcement learning in continuous action space