Similar Items: The ODE Method for Stochastic Approximation and Reinforcement Learning with Markovian Noise