Researchers at the Ohio State University Introduce Famba-V: A Cross-Layer Token Fusion Technique that Enhances the Training Efficiency of Vision Mamba Models