Biological Neurons Compete with Deep Reinforcement Learning in Sample Efficiency in a Simulated Gameworld