Poor performance for QC filtering on medium sized genotype data

[tpoterba]

Should also mention, SIMD in Hail is definitely on the roadmap. We do also have a lot of performance gains on the table in the JVM code we generate, though.