24. Cross-Attention Multi-Scale Vision Transformer for Image Classification (CrossViT)