🦀 SOC projections in Rust and other improvements

[alphaville]

Main Changes

Summary: Gradgen v0.5.1 brings second-order cones (SOCs), which are useful in convex optimization. It also brings significant improvements in the generated code leading to a more "economic" allocation of workspace memory. Earlier, workspace would almost always be required, whereas now we check whether the output variable can serve as workspace. More thorough unit testing is also done (we now have 537 unit tests). This includes integration tests where we compare the results obtained by gradgen to sympy.

• SOC projections and squared distances
• Workspace variables are not generated unless required
• Unit tests are in debug mode now
• A few bugs fixed (related to code generation)
• Further integration tests (comparisons with sympy)

Associated Issues

• Closes Projection on SOC and squared distance function #78
• Closes Superfluous workspace memory allocation #81

Merge Dependencies

• None (but, better merge 📝 Docusaurus #77 first)

Target versions and releases

• Gradgen v0.5.1
• Alpha release v0.5.1a2

Checklist

• API documentation updated
• Appropriate testing
• Create a demo in demos/, if necessary
• Update CHANGELOG
• Update webpage documentation (incl. Google Colab demo), if necessary
• Bump versions in pyproject.toml, if necessary
• If a new version is being released, write blog post