Reinforcement Learning and Feedback Control for High-Level Upper-Extremity Neuroprostheses