Tri-Perspective View for Vision-Based 3D Semantic Occupancy Prediction