Fixed exception in SplitNCigarReads for non-reference consuming reads

[jamesemery]

Fixes #5293

[lbergelson]

Looks good to me. My only thought is that it might make more sense to handle these with an upfront filter because they don't really make any sense.

[codecov-io]

Codecov Report

Merging #5298 into master will increase coverage by <.001%.
The diff coverage is 90.909%.

@@               Coverage Diff               @@
##              master     #5298       +/-   ##
===============================================
+ Coverage     86.777%   86.777%   +<.001%     
- Complexity     29910     29912        +2     
===============================================
  Files           1835      1835               
  Lines         138574    138583        +9     
  Branches       15255     15256        +1     
===============================================
+ Hits          120250    120258        +8     
- Misses         12772     12774        +2     
+ Partials        5552      5551        -1

Impacted Files	Coverage Δ	Complexity Δ
...er/tools/walkers/rnaseq/OverhangFixingManager.java	94.304% <100%> (+0.073%)	70 <0> (ø)	⬇️
.../walkers/rnaseq/OverhangFixingManagerUnitTest.java	97.838% <85.714%> (-0.477%)	33 <1> (+1)
...utils/smithwaterman/SmithWatermanIntelAligner.java	50% <0%> (-30%)	1% <0%> (-2%)
...ithwaterman/SmithWatermanIntelAlignerUnitTest.java	60% <0%> (ø)	2% <0%> (ø)	⬇️
...roadinstitute/hellbender/utils/read/ReadUtils.java	79.812% <0%> (+0.704%)	201% <0%> (+2%)	⬆️

[jamesemery]

@lbergelson It may, but I want them to get passed through the splitter and the filter and be written on the other side, which requires they get emitted into the splitter. I would rather the tool no remove reads from someones bam as maybe there is some reason their aligner emitted these reads, hence we emit them and add a check that prevents the bogus genomeLocs from being created for them (so they don't get processed by the manager).