Improve handling haploid calls

[VorontsovIE]

Here are four fixes which should improve handling haploid genotypes:

• Function revise_genotypes now skips haploid calls. Earlier it was failing due to incorrect number of genotypes (n_allele haploid genotypes instead of n_allele*(n_allele+1)/2 diploid genotypes).
• Introduce new value NoCallReason::HaploidCall to mark alleles missing because they were missing in original haploid call. This reason is denoted by letter 'H' in RNC field.
• FormatHelpers for Number=G[enotype] fields just skip and censor out samples which has haploid records for a site.
• Haploid calls get genotypes A/. instead of ./A.

This patch should fix #162.

[coveralls]

Coverage decreased (-0.09%) to 93.315% when pulling 47e9465a714cdfa3057a4cc415546943693617a7 on VorontsovIE:fix_haploid_calls into ca0e9b774c8006faa5c3d7659f89653bbbc92d6b on dnanexus-rnd:master.

[mlin]

@VorontsovIE my apologies for the slow uptake on this. It looks just right to me, but I'd like to add a new gVCF test case to lock it down against future changes. I do still have those DRAGEN gVCFs for CEU trio, but I need to get to it.........has this diff unblocked you for the time being?

Also, I really appreciate how you were able to dig into the different parts of the genotyper code. You might have seen we're considering important changes to how it works in the medium-term future, specifically around multiallelic sites. Really eager for community feedback on that from any interested parties!