feat: add vectorized 3D cartesian utils and polar pytests

[Riteesh-NITT]

Describe your changes

Summary of the changes:
This PR introduces vectorized PyTorch utilities (spherical_utils.py) to convert 2D geographic coordinates (latitude/longitude) into 3D Cartesian coordinates $(x, y, z)$. It also includes a pytest suite specifically validating mathematical stability at the North Pole singularity and across the anti-meridian (Date Line).

Note: Hi @leifdenby and @mandeepsingh2007, following up on our thread in #76/#77! I know we are waiting on the load_graph shim (Issue #385) to be merged before fully hooking this up, but I wanted to get these utilities up for early review. Let me know if the tensor shapes align with the native PyG pipeline you are building. I'll keep this as a Draft until the base branch is ready!

Motivation and context:
When mapping 2D weather grid data to a 3D icosahedral mesh, flat lat-lon grids suffer from severe spatial distortion at the poles. This utility provides the mathematical bridge to project grid nodes onto a unit sphere in 3D space, which is an architectural prerequisite for accurate Grid-to-Mesh (g2m) distance calculations and connectivity.

Dependencies:

• torch
• pytest (for the test suite)

Issue Link

Addresses #76 and #77

Type of change

• 🐛 Bug fix (non-breaking change that fixes an issue)
• ✨ New feature (non-breaking change that adds functionality)
• 💥 Breaking change (fix or feature that would cause existing functionality to not work as expected)
• 📖 Documentation (Addition or improvements to documentation)

Checklist before requesting a review

• My branch is up-to-date with the target branch - if not update your fork with the changes from the target branch (use pull with --rebase option if possible).
• I have performed a self-review of my code
• For any new/modified functions/classes I have added docstrings that clearly describe its purpose, expected inputs and returned values
• I have placed in-line comments to clarify the intent of any hard-to-understand passages of my code
• I have updated the documentation to cover introduced code changes
• I have added tests that prove my fix is effective or that my feature works
• I have given the PR a name that clearly describes the change, written in imperative form (context).
• I have requested a reviewer and an assignee (assignee is responsible for merging)

[Riteesh-NITT]

Hi @leifdenby and @joeloskarsson, just putting this up as a Draft! This implements the vectorized spherical-to-cartesian conversions and the polar edge-case tests we discussed in #76 and #77. Hey @mandeepsingh2007, here is the math bridge for the icosahedral mesh generation. Feel free to pull this branch locally if you need to test the grid-to-mesh (g2m) mappings with your trimesh output!

[mandeepsingh2007]

Hey @Riteesh-NITT , I just pulled the logic locally and ran some integration tests against the current NumPy g2m geometry we built in PR #76.

The outputs from your torch implementation match the current numpy pipeline outputs down to the exact floating-point precision (max difference ~1e-8).

This confirms that your PyTorch vectorization is a perfect drop-in replacement. Once we move to the HeteroData native graph loading, we can safely swap the NumPy processing out for this without breaking any downstream spatial alignments.

[leifdenby]

since this functionality is a utility being added for icosahedral meshes, which in turn was decide in today's dev meeting has gone on the roadmap for v0.5.0, I will put this utility on the roadmap for v0.5.0 as well. Although I would prefer to just have this addition made with the functionality to use it rather than having a util that isn't used yet