Q-PETR: Quant-aware Position Embedding Transformation for Multi-View 3D Object Detection