Added support for importing dosages as call values

[tmwong2003]

This PR adds support for optionally loading a dosage values as the call value for genotype loci. This is to support VCF files (such as the ones we use at 23andMe) that have dosages (DS) instead of genotype call information (typically GT).

[tpoterba]

I think the ability to load entry floats as float32 is a great change, but I think all the language about "dosage" is a bit use-case-specific and misleading. My proposal would be to add a parameter to import_vcf called entry_float_type or something, which defaults to hl.tfloat64.

This would get passed in as a config param, and would let us do the following in the headerField calculation:

    case (VCFHeaderLineType.Float, false) => entryFloatType

I don't think this will be a huge change from what you have, and will mostly involve deleting code.

looking again

[tpoterba]

a few more small changes, and good to go.

[tpoterba]

looks like a compile error in TestUtils.scala.

The rest looks good, will approve when tests pass

addressed

[tmwong2003]

looks like a compile error in TestUtils.scala.

The rest looks good, will approve when tests pass

I fixed TestUtils.scala; the issue was a missing parameter in a call to the (changed) MatrixVCFReader.

Should I rebase and squash all of my commits into a single commit, or are you OK with merging my branch with the discrete feature commits (i.e., non-"Merge remote-tracking branch" commits)?

[tpoterba]

looks good to me!