Add non-coding regions

[donkirkby]

Add some coordinate sequences that are not coding for proteins. The sequences can be marked as nucleotide sequences.

• Add missing regions for SARS-CoV-2.
• Include them in nuc.csv, but not amino.csv.
• Align amino sequences.
• Only count overlapping minimap matches once.
• Report partial codons at boundary of minimap match.
• Moved to separate issue: Add coverage plots and scores for non-coding regions #725
• Add missing regions for HIV.
• Stop trimming stop codons from count files.

[donkirkby]

The plan is to first use minimap2 to align the entire consensus sequence with the whole-genome nucleotide reference, then cut out the sections that align to each gene region, and use the current Gotoh technique to align each of the three reading frames and choose the best match. That lets us keep the benefit of aligning in amino space, but gain the benefit of minimap2 handling the huge deletions we see in the assembled proviral samples. The sections that minimap2 can't align will appear in the genome coverage plots, but not in any gene coverage plots. Coordinate regions that are nucleotide references will just use the minimap2 alignment, without the extra Gotoh step.
Question: what do we do with minimap2 alignments that overlap at the ends? For the genome coverage plot, I just let the alignment closer to the 5' end override the other one. The overlaps are usually small enough that it's not noticable on the plot. In the amino.csv and nuc.csv reports, though, you can see the details.
I'm moving this issue to the next release, because it's going to take a while to finish.

[donkirkby]

I discussed the problem of overlaps and gaps between minimap2 matches with @cbrumme, and we agreed to use the same arbitrary rule that I currently use in the plots. Any gaps between matches will be labelled as yellow in the coverage plots, and won't be reported in any gene region.