Main Changes

I wanted to try the core crate out on a microcontroller that just has hard-floats for f32, so I hacked this together. I'm only using PANOC without constraints, so I'm not sure if anything really works that well outside of what I used. Perhaps the complexity is somewhat inevitable to support f32/f64, but it's annoying how much extra boilerplate is added with this. Nonetheless, it could be a good basis for adding f32 support here.

• Make a OptFloat trait
• Provide PANOC constants implemented for f32 and f64
• Use somewhat arbitrary values for f32 PANOC constants, unclear how good these are. These should be looked at.
• Add some tests for f32 (some of which are failing!)

Associated Issues

• Related to single precision (float32, f32) support is missing #372

TODOs

• Documentation
• All tests must pass
• Update CHANGELOG(s)
• Update webpage documentation
• Bump versions (in CHANGELOG, Cargo.toml and VERSION)
• Use non-forked version of lbfgs-rs