Tempo Adaptation in Non-stationary Reinforcement Learning