Add SLIM/fsSLIM scorer

[mdekstrand]

This PR adds the SLIM and fsSLIM models for scoring. Closes #894.

The implementation has a similar structure to the original by the Karypis lab's, but stores residuals instead of estimates for driving the coordinate descent updates. This makes the inner optimization step a little simpler.

It is very different from the RecBole implementation, which uses SciKit-Learn's elastic net implementation to learn item columns.

The Regularization Paths paper specifies that the predictors are standardized, but the Karypis code does not seem to center the predictors, only scale them (implicitly by scaling the updates). The RP paper says it is “obvious” how to extend their method to sparse data like this, but do not provide or explain the relevant derivations. However, if Karypis didn't center, I think we can get away with not centering.

The way scaling is implemented does raise some questions about numerical stability, but I'm not sure that adjusting how the scaling works would fix that concern, or even if it's really a problem.

References:

• SLIM
• SLIM reference implementation
• Regularization Paths
• RustyStats coordinate descent

[codecov]

Codecov Report

❌ Patch coverage is 97.89916% with 5 lines in your changes missing coverage. Please review.
✅ Project coverage is 89.52%. Comparing base (c47d140) to head (6ba6aa4).
⚠️ Report is 37 commits behind head on main.

Files with missing lines	Patch %	Lines
src/accel/slim/mod.rs	97.07%	5 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #896      +/-   ##
==========================================
+ Coverage   89.38%   89.52%   +0.13%     
==========================================
  Files         226      228       +2     
  Lines       15607    15845     +238     
==========================================
+ Hits        13951    14185     +234     
- Misses       1656     1660       +4     

☔ 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.