mdp_Q_learning function

Solves discounted MDP using the Q-learning algorithm (Reinforcement Learning)