SemParamsPenalty loss term for flexible regularization

[alyst]

• single term to handle ridge/lasso/elasticnet and more
• support for hinge penalties
• allow l^alpha for any alpha
• allow affine transform of parameters before regularization
• SemSpec rather then SemImplied in ctor

[codecov]

Codecov Report

❌ Patch coverage is 86.18421% with 21 lines in your changes missing coverage. Please review.
✅ Project coverage is 73.34%. Comparing base (1f8d2a9) to head (7a485d3).
⚠️ Report is 122 commits behind head on devel.

Files with missing lines	Patch %	Lines
src/additional_functions/sparse_utils.jl	53.84%	18 Missing ⚠️
src/loss/regularization/params_penalty.jl	97.16%	3 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##            devel     #328      +/-   ##
==========================================
+ Coverage   71.83%   73.34%   +1.50%     
==========================================
  Files          51       58       +7     
  Lines        2223     2604     +381     
==========================================
+ Hits         1597     1910     +313     
- Misses        626      694      +68     

☔ View full report in Codecov by Harness.
📢 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.

[alyst]

@Maximilian-Stefan-Ernst It should be ready for review

[alyst]

@Maximilian-Stefan-Ernst I've added some convenience functions for working with parameters and model-derived sparse matrices that I have forgotten to cherry-pick.
I have also updated the regularization documentation with the new regularization terms.

[alyst]

@Maximilian-Stefan-Ernst With v0.5 released, this refactoring/extension of regularization terms could, in principle, be merged into dev, if you think it's worth it.