Moar significant figures for percentages

[eparejatobes]

The taxa frequencies are rounded to zero in more cases than what would be reasonable to expect, leading to values that don't add up to 100% by a significant margin.

A good rule of thumb could be for the frequency percentage be precise enough to detect one read in a sample; we can take 10^6 as an upper bound for the number of reads per sample.

Reported by @rtobes

[rtobes]

I had thought the same.

[laughedelic]

Ok, but I can do it only for frequencies, not for pident, because it comes out of BLAST with maximum 2 digits. Anyway, it shouldn't be a problem, because we have only high values of pident.

[laughedelic]

🚢 done