Reinforcement Learning with Neural Radiance Fields