Validate HCV subtype calls

Some of the HCV resistance rules that we wanted to add in #436 depended on treating samples differently based on their subtype. However, we're not very confident in the MiCall pipeline's current subtype calls. We're worried that the remap drift we saw in issue #290 could still happen between subtypes.

Here's the proposed validation process:

• collect a large set (10 per subtype?) of HCV reference sequences with subtype annotations (maybe from the mixed-hcv pipeline) (I ended up using the web reference collection from hcv.lanl.gov, as shown in the script.)
• build a phylogenetic tree to validate those subtypes (hopefully with help from @greenbeard1 ) (Genotype 6 is paraphyletic - close enough. Others are all monophyletic.)
• look through all of the samples that MiCall has called a subtype on, and select up to 100 from each subtype, then measure the Levenshtein distance from each sample consensus to all the reference sequences, and finally report any samples that were called wrong (Ended up using all the samples I validated release 7.9 with. They all ran in 1hr 53min.)
• see if it's helpful to add more references to MiCall pipeline from web reference collection: particularly 3g and 3h
• compare FastTree results with Levenshtein distance comparison (Levenshtein wasn't an improvement.)
• compare FastTree results with BlastN searches
• if there are subtypes that we've never processed, we could try simulating data based on the references

If we find a lot of bad calls, we might need to switch to de novo assembly, as described in #442.

Start with the script used for the web reference collection and the existing FastTree and Levenshtein validation results. Compare the remaining FastTree and BlastN checks, assess whether references such as 3g and 3h are needed, and document any incorrect subtype calls or recommended pipeline changes.