Multi-Manifold Attention for Vision Transformers