feat: add spherical coordinate support via coords_crs parameter

[opeblow]

Describe your changes

This PR adds spherical coordinate support to weather-model-graphs,
enabling the library to work with lat/lon coordinates (EPSG:4326)
instead of only flat Cartesian coordinates.

Changes made in src/weather_model_graphs/create/base.py:

• Added coords_crs parameter to accept input coordinate reference system
• Added graph_crs parameter to specify projection for graph construction
• When coords_crs is provided, coordinates are automatically projected
  from lat/lon degrees to a metre-based CRS before graph construction
• Fixed bug where max_num_neighbours=1 caused a TypeError because
  numpy returned a scalar instead of an array — fixed with np.atleast_1d()

Added tests in tests/test_spherical_coords.py:

• Test 1: Flat Cartesian coordinates still work (backwards compatible)
• Test 2: Spherical/geographic coordinates work with real metre-based
  edge distances (e.g. 394,581m, 789,163m) proving haversine projection

Issue Link

closes #42

Type of change

• Bug fix (non-breaking change that fixes an issue)
• New feature (non-breaking change that adds functionality)

[opeblow]

Hello 👋🏿🤗

[opeblow]

Hi, just checking in on this PR.
Happy to make any changes if needed.
Let me know if there's anything
blocking the review.

[sadamov]

Thanks for the contribution @opeblow. Closing this in favour of existing PRs that cover the same ground:

The np.atleast_1d fix for max_num_neighbours=1 (#83) is already addressed in #90 and #113.
Orphan/disconnected node detection is covered more thoroughly by #132, which also handles geometric and hierarchical consistency checks.
The PR description (spherical coordinate support) also doesn't match the actual diff - coords_crs/graph_crs have been in main for a while already.

If you want to stay involved, #90 has a change request from @leifdenby that could use a follow-up, or #132 could use review attention.