Reinforcement Learning for F1tenth Racing: Bridge the Sim-to-Real Gap