End coordinate should not be compared when sorting PE evidence

[mwalker174]

• Use start position as end for purposes of sorting PE evidence
• Adds regression test to MultiFeatureWalker

[tedsharpe]

Always thought that the definition of getEnd() was totally weird. This is much more sensible.
I notice that the little test file that you added has no instances of startContig==endContig, so it won't actually test the change you made to the definition of getEnd(), but otherwise this looks fine.
Sorry that I didn't anticipate the bug.

[codecov]

Codecov Report

Merging #7835 (20f2ef8) into master (e0e6669) will increase coverage by 0.000%.
The diff coverage is 92.857%.

@@             Coverage Diff             @@
##              master     #7835   +/-   ##
===========================================
  Coverage     86.954%   86.954%           
- Complexity     36897     36898    +1     
===========================================
  Files           2214      2214           
  Lines         173540    173551   +11     
  Branches       18736     18738    +2     
===========================================
+ Hits          150900    150910   +10     
- Misses         16037     16039    +2     
+ Partials        6603      6602    -1     

Impacted Files	Coverage Δ
...ples/ExampleMultiFeatureWalkerIntegrationTest.java	91.667% <92.308%> (+0.100%)	⬆️
...te/hellbender/tools/sv/DiscordantPairEvidence.java	73.333% <100.000%> (-1.667%)	⬇️
...titute/hellbender/tools/walkers/sv/SVAnnotate.java	98.701% <0.000%> (ø)
...itute/hellbender/tools/LocalAssemblerUnitTest.java	92.448% <0.000%> (ø)