MATE: Multi-view Attention for Table Transformer Efficiency