Add record level reporting and PyTorch support to StructuralAttack

[kayatefi]

• Add store record level
• Support PyTorch models

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 99.42%. Comparing base (3acfd9e) to head (3ed3916).
⚠️ Report is 1 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #367      +/-   ##
==========================================
- Coverage   99.67%   99.42%   -0.26%     
==========================================
  Files          44       23      -21     
  Lines        4363     2615    -1748     
==========================================
- Hits         4349     2600    -1749     
- Misses         14       15       +1     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[jim-smith]

@kayatefi Before we can merge this pull request we should define a simple case where we know what the correct record-level output should be and then write an appropriate test to see it is generated properly.

I was thinking of doing something simple with a decision tree and a small dataset- maybe 20 cases (some being duplicates), of 2/3 variables with an xor decision function.

• scenario 1: decision tree max depth 1: should get 50% accuracy, at least 10 records in each leaf so 5 of each class ==> should fail all the strucutural tests
• scenario 2: let the tree grow deeper: should her able to manipulate it to show class disclosure and small group problems

[jim-smith]

@rpreen I've been through Kay's code which was mostly fine then added the individual record-level reporting, which meant I had to rewrite/refactor it a bit.

Everything should be tested for correct functionality as well as coverage.