We are not likely to add this option in QIIME 2, as it can introduce some serious issues downstream, e.g., for taxonomic classification. Basically, just concatenating non-overlapping pairs end-to-end can mess up alignments (e.g., for phylogeny building) and kmer profiles (e.g., used for taxonomy classification) because these concatenated reads are artificial constructs. You can read more about these concerns here, also a number of other forum topics that discuss this option: