A Zero-Shot Reinforcement Learning Strategy for Autonomous Guidewire Navigation