A General Control-Theoretic Approach for Reinforcement Learning: Theory and Algorithms