Accelerating Quantum Reinforcement Learning with a Quantum Natural Policy Gradient Based Approach