Bridging the Gap Between Offline and Online Reinforcement Learning Evaluation Methodologies