Change default blocking and NON-REF LOD params for M2 GVCF mode

[ldgauthier]

Teeny tiny PR -- not happy with the AF thresholding for GVCF mode yet

[ldgauthier]

@droazen I'd like this to squeeze into 4.1. Tests pass locally after the second commit.

[codecov-io]

Codecov Report

Merging #5615 into master will decrease coverage by 0.151%.
The diff coverage is 100%.

@@               Coverage Diff               @@
##              master     #5615       +/-   ##
===============================================
- Coverage     87.037%   86.886%   -0.151%     
- Complexity     31537     31614       +77     
===============================================
  Files           1930      1933        +3     
  Lines         145455    145939      +484     
  Branches       16090     16178       +88     
===============================================
+ Hits          126600    126801      +201     
- Misses         12996     13255      +259     
- Partials        5859      5883       +24

Impacted Files	Coverage Δ	Complexity Δ
...r/tools/walkers/mutect/Mutect2IntegrationTest.java	91.623% <ø> (ø)	89 <0> (ø)	⬇️
...der/tools/walkers/mutect/M2ArgumentCollection.java	88.889% <100%> (ø)	14 <0> (ø)	⬇️
...ithwaterman/SmithWatermanIntelAlignerUnitTest.java	60% <0%> (-30%)	2% <0%> (ø)
...oadinstitute/hellbender/utils/gcs/BucketUtils.java	64.961% <0%> (-13.271%)	56% <0%> (+17%)
...rg/broadinstitute/hellbender/utils/io/IOUtils.java	59.877% <0%> (-13.105%)	114% <0%> (+19%)
...utils/smithwaterman/SmithWatermanIntelAligner.java	80% <0%> (-10%)	3% <0%> (ø)
...broadinstitute/hellbender/engine/FeatureInput.java	86.869% <0%> (-6.187%)	26% <0%> (+8%)
...institute/hellbender/engine/FeatureDataSource.java	74.556% <0%> (-1.914%)	54% <0%> (+12%)
...lotypecaller/readthreading/ReadThreadingGraph.java	88.608% <0%> (-0.253%)	144% <0%> (-1%)
...notyper/GenotypeCalculationArgumentCollection.java	100% <0%> (ø)	4% <0%> (+2%)	⬆️
... and 5 more

[droazen]

@davidbenjamin please review

[davidbenjamin]

You can take or leave the comment about avoiding the initialization block. 👍

[davidbenjamin]

Could you use GVCFGQBands = Arrays.asList(MathUtils.createEvenlySpacedPoints(-2.5, 1, 8))?

[ldgauthier]

I'm going to keep the initialization as is because it mirrors HaplotypeCaller and because I feel like defining blocks by their size is more intuitive than defining the number of blocks between two bounds.