VLM-E2E: Enhancing End-to-End Autonomous Driving with Multimodal Driver Attention Fusion