Reinforcement Learning And Feedback Control For High-Level Upper-Extremity Neuroprostheses