Safety Implications of Explainable Artificial Intelligence in End-to-End Autonomous Driving