Tri-Level-ViT. [AAAI 2023 Oral] Peeling the Onion: Hierarchical Reduction of Data Redundancy for Efficient Vision Transformer Training