White-Box Transformers via Sparse Rate Reduction: Compression Is All There Is?