Triplet Attention: Rethinking the Similarity in Transformers