Multiagent reinforcement learning: algorithm converging to Nash equilibrium in general-sum discounted stochastic games