Steady State Analysis of Episodic Reinforcement Learning