Make sure the GATK tool python package is present before executing streaming commands.

[cmnbroad]

Fixes #5751 and #4591. Longer term we'll still want to do package version-checking/verification per #4995 as well.

@jamesemery I included tests for this change, but I need the tests to only run when the conda env is NOT activated. Unit tests are always run on the docker image, so thats out. Integration tests are run on both the docker and the travis image, so I throw a skip exception on the docker, which I detect using the "CI" env variable. But that seems fragile and confusing. Is there a better way to do this ?

[jamesemery]

@cmnbroad I see. The "CI" variable does seem brittle, especially since I'm not strictly sure where it is set. I think a somewhat safer place would be to add some global test flag to the docker image would be to add it to the run_unit_tests.sh script. That way we know it is getting triggered exactly before we run the tests in just the docker environment. Is there some way of detecting what conda environment is active outside of conda.

[codecov-io]

Codecov Report

Merging #5819 into master will decrease coverage by 6.733%.
The diff coverage is 41.667%.

@@               Coverage Diff               @@
##              master     #5819       +/-   ##
===============================================
- Coverage     87.031%   80.298%   -6.733%     
+ Complexity     32107     30526     -1581     
===============================================
  Files           1972      1975        +3     
  Lines         147194    147487      +293     
  Branches       16201     16233       +32     
===============================================
- Hits          128104    118429     -9675     
- Misses         13184     23343    +10159     
+ Partials        5906      5715      -191

Impacted Files	Coverage Δ	Complexity Δ
.../walkers/vqsr/CNNScoreVariantsIntegrationTest.java	97.386% <0%> (-2.614%)	13 <0> (ø)
...er/utils/python/StreamingPythonScriptExecutor.java	85.47% <100%> (+0.125%)	22 <1> (ø)	⬇️
...va/org/broadinstitute/hellbender/GATKBaseTest.java	96.429% <100%> (-3.571%)	7 <2> (ø)
...python/StreamingPythonExecutorIntegrationTest.java	11.111% <11.111%> (ø)	1 <1> (?)
.../hellbender/utils/python/PythonScriptExecutor.java	66.667% <66.667%> (+3.03%)	10 <0> (ø)	⬇️
...kers/filters/VariantFiltrationIntegrationTest.java	0.826% <0%> (-99.174%)	1% <0%> (-25%)
...dorientation/CollectF1R2CountsIntegrationTest.java	0.917% <0%> (-99.083%)	1% <0%> (-12%)
.../walkers/bqsr/BaseRecalibratorIntegrationTest.java	1.031% <0%> (-98.969%)	1% <0%> (-7%)
...s/variantutils/VariantsToTableIntegrationTest.java	1.042% <0%> (-98.958%)	1% <0%> (-20%)
...ers/vqsr/FilterVariantTranchesIntegrationTest.java	1.053% <0%> (-98.947%)	1% <0%> (-5%)
... and 180 more

[droazen]

@jamesemery @cmnbroad Can we try to get this one in in time for the release?

[jamesemery]

This looks fine to me if you have empirically observed the tests (and that you have seen that they run on the non-docker image travis machine).

[jamesemery]

I still maintain that we should change the "CI" variable and set it ourselves as part of the run-docker script so we are not relying on some feature of the environment that might change. Otherwise i'm not too bothered by this being a skip exception.

[jamesemery]

ehh... this could be extracted out into a skipIfNotInDockerImage() since it's copied in two places.

[jamesemery]

These changes look good, feel free to merge when tests pass 👍