Rust moments

[arthur-lin1027]

An updated version of @YCC-ProjBackups's PR. The code has changed significantly since last year so it's annoying to rebase.

• I've added code that builds the rust source code in a single pip install .
  • Removed some relic setup stuff (deleted setup.py and setup.cfg) in favor of a single pyproject.toml file that has been properly configured to compile rust code.
  • This improves upon Yong's version where the use of relative imports made building the FFI difficult.
  • I've also added some timing tests that indeed show the rust port to be much quicker, anywhere between 30-150x faster.

One issue is still with the hand-created determinant code, which fails to reproduce the numpy linear algebra code. The workaround isn't too straightforward -- this is either a skill-issue for me, or numpy-rust being immature. I wasn't able to get numpy-rust to interface with nalgebra, the main rust linear algebra package.

• Supposedly the function PyReadonlyArray::try_as_matrix exists to convert types from numpy-rust to nalgebra, but I couldn't get this to work.
• For now, the determinants only diverge for near-singular matrices. We calculate determinants of our dilation matrices, which for standard use cases should never be close to singular.
  • The dilation matrix should have a determinant that is almost equal to the inverse product of our semiaxes lengths (for GTOs, the determinant is distorted a bit by the GTO width)

[rosecers]

There are still some things that need to be cleaned up, and I'd suggest deleting the rust folder if the file inside it is no-longer necessary. That was causing some confusion.

But overall looks great!