Text this: AlignFusionNet: Efficient Cross-Modal Alignment and Fusion for 3D Semantic Occupancy Prediction