DAT++: Spatially Dynamic Vision Transformer with Deformable Attention