A Unifying View of Optimism in Episodic Reinforcement Learning