New kernel implementations, nominal model implementation, and training CLI

[laserkelvin]

Proposed changes

This PR is an omnibus of changes associated with the recently accepted paper at AI4Mat - NeurIPS 2024.

Types of changes

• Bugfix (non-breaking change which fixes an issue)
• New feature (non-breaking change which adds functionality)
• Breaking change (fix or feature that would cause existing functionality to not work as expected)
• Documentation update (if none of the other choices apply)

Checklist

Put an x in the boxes that apply. You can also fill these out after creating
the PR. If you are unsure about any of them, do not hesitate to ask. We are
here to help! This is simply a reminder of what we are going to look for before
merging your code.

• I have read the CONTRIBUTING agreement
• Current formatting and unit tests / base functionality passes locally with my changes
• I have added tests that prove my fix is effective or that my feature works (if appropriate)
• I have added necessary documentation (if appropriate)
• Any dependent changes have been merged and published in downstream modules

Further comments

The main things to highlight in this PR are:

• New Jupyter notebooks that reproduce the new spherical harmonic derivations, which have been refactored so that they are direct functions of $xyz$.
• New Triton lang implementations of these kernels, contained in the submodule equitriton.sph_harm.direct, including the relevant torch.autograd.Function bindings and a optional functional interface.
• A nominal model implementation that uses the new kernels, along with a PyTorch Lightning workflow for training and testing against QM9. The model implementation is contained in the equitriton.models submodules, and the training scripts plus the PHATE projection analysis are in scripts.
• Updated the README with a nominal paper citation, and guidance on the new kernels.