feat: add spherical geometry utilities for global domain support

[arzoo0511]

Describe your changes

This PR introduces basic spherical geometry utilities to support future global forecasting capabilities in Neural-LAM.

Summary of changes

• Added lat_lon_to_cartesian() for converting latitude–longitude coordinates to Cartesian coordinates on the unit sphere
• Added get_area_weights() for computing latitude-based area weighting
• (Optional) Added helper method in datastore to expose Cartesian coordinates

Motivation and context

Current implementations assume planar geometry, which does not extend well to global forecasting. These utilities provide a reusable and modular foundation for geometry-aware graph construction on a spherical domain, aligning with ongoing work toward global support.

Dependencies

• No additional dependencies introduced
• Uses existing PyTorch utilities

---

Issue Link

Related to global forecasting discussions (no specific issue linked)

---

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
• 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 README (not required for this change)
• I have added tests (planned in follow-up PR)
• I have given the PR a name that clearly describes the change, written in imperative form

---

Notes

This PR is intentionally scoped to be small and modular, focusing only on geometry utilities. Future work will build on this to implement global graph construction and integration into the training pipeline.

[Copilot]

Pull request overview

Adds initial spherical-geometry utilities intended to support future global-domain workflows in Neural-LAM.

Changes:

• Introduces lat_lon_to_cartesian() for converting latitude/longitude (degrees) to 3D unit-sphere Cartesian coordinates (PyTorch).
• Introduces get_area_weights() for latitude-based area weighting.
• Adds a helper get_cartesian_coords (currently implemented as a free function).

Reviewed changes

Copilot reviewed 2 out of 3 changed files in this pull request and generated 6 comments.

File	Description
geometry.py	Adds spherical-geometry helper functions (lat/lon → Cartesian, latitude area weights, and a Cartesian-coords helper).
.cora/analytics/index.json	Adds a generated analytics index artifact.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[sadamov]

#473 is currently a draft PR that addresses these concerns already. it would be great if you could hop over there and contribute to that PR/discussion directly.