Performative Reinforcement Learning in Gradually Shifting Environments