A Coding Implementation to Train Safety-Critical Reinforcement Learning Agents Offline Using Conservative Q-Learning with d3rlpy and Fixed Historical Data