Fix indices in LOCO for record-level insights and add more robust tests

[Jauntbox]

Related issues
N/A

Describe the proposed solution
Simple fix to the indexing of insights that didn't show up because the tests were too simple. Adds a significantly more complex test that includes a more realistic feature engineering DAG, and tests a variety of properties that should hold for record-level insights for randomly generated datasets with a strong relation between the features and the label.

Describe alternatives you've considered
N/A

Additional context
N/A

[codecov]

Codecov Report

Merging #216 into master will increase coverage by 0.01%.
The diff coverage is 100%.

@@            Coverage Diff             @@
##           master     #216      +/-   ##
==========================================
+ Coverage   86.34%   86.36%   +0.01%     
==========================================
  Files         310      310              
  Lines       10137    10137              
  Branches      548      542       -6     
==========================================
+ Hits         8753     8755       +2     
+ Misses       1384     1382       -2

Impacted Files	Coverage Δ
...e/op/stages/impl/insights/RecordInsightsLOCO.scala	94.11% <100%> (ø)	⬆️
...es/src/main/scala/com/salesforce/op/OpParams.scala	89.79% <0%> (+4.08%)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update cd2fa5b...9dd893f. Read the comment docs.

[tovbinm]

Lgtm, the test is bizarre. Perhaps use withClue around asserts to provide more details in case of failures?