gather behavior around identical sketch matches is confusing and ~error-prone

[ctb]

The root question that led to discovering #2318 was that @bluegenes was getting unexpectedly inaccurate results from a zymo mock community analysis, and she wanted to understand why.

In brief, there were two genomes that were producing identical sketches, GCA_902364275.1 Faecalibacterium prausnitzii, MGYG-HGUT-00195 and GCA_003478405.1 Faecalibacterium sp. AF28-13AC strain=AF28-13AC. And of course sourmash prefetch was finding both of them, and gather was picking one of them.

If it picked the first, the species was "correctly" assigned as Faecalibacterium prausnitzii because that's the NCBI taxonomy entry for this genome, and the mock community composition indicated it contained F. prausnitzii. If it picked the second, then it was missing the correct species.

See #1615 and the idea for making identical matches accessible. It is not clear to me how to properly handle this in gather output, since sourmash tax only sees gather output. Maybe we could provide multiple entries per rank where appropriate, and mark them as "secondary", and then sourmash tax could choose the one with the most specific tax, or something? Ugh.