An Efficient Transformer for Multi-Task Policy Learning