Fix bug where single-base intervals led to no calls

[davidbenjamin]

Closes #6495. @ldgauthier do you want this one?

[gatk-bot]

Travis reported job failures from build 29539
Failures in the following jobs:

Test Type	JDK	Job ID	Logs
R	openjdk8	29539.6	logs
python	openjdk8	29539.5	logs
variantcalling	openjdk8	29539.4	logs
unit	openjdk8	29539.3	logs
integration	openjdk8	29539.2	logs

[gatk-bot]

Travis reported job failures from build 29541
Failures in the following jobs:

Test Type	JDK	Job ID	Logs
variantcalling	openjdk8	29541.4	logs

[ldgauthier]

This looks good to me, pending tests passing. I find it a little disconcerting that we apparently were skipping a mitochondrial variant in the M2 tests, but at least now we have a regression test for that.

[davidbenjamin]

I find it a little disconcerting that we apparently were skipping a mitochondrial variant in the M2 tests

Nothing to worry about — that number change (if that's what you're referring to) is just a hard-coded read count. It changes every time windows are adjusted.

[ldgauthier]

Oh, good, I thought it was the variant position!