Split reference sequences into different fasta files by their annotation from `treesapp purity`

[cmorganl]

treesapp purity is (mostly) good at indicating whether reference packages will end up classifying off-target homologs. A missing piece, however, leaves users unsure of what to do with the reference package if off-target hits were found.

While not perfect, the sequences of a reference package could be split across fasta files based on what orthologous group they were classified to. A fasta containing all sequences that were not classified would be written as well. With these files, users could concatenate the sequences they believe to belong to their targeted protein family and recreate the refpkg.

This would be ideal in cases when misannotated, nonhomologous sequences were included in the initial set used by treesapp create.