Warn on degree/meter unit mismatch in planar aspect()

[brendancol]

Closes #2839

What this does

• aspect()'s planar path now calls warn_if_unit_mismatch(agg) before computing resolution, the same guard slope() already has. Degree-based lat/lon rasters with meter elevations get a UserWarning instead of silently producing wrong values.
• Adds a "Notes" paragraph to the aspect() docstring mirroring slope's, explaining the warning and pointing to method='geodesic' or reprojection.

Backend coverage

The warning lives in the shared aspect() entry point ahead of backend dispatch, so it fires for all four backends (numpy / cupy / dask+numpy / dask+cupy). No backend kernels changed.

Test plan

• test_planar_warns_on_degree_coords_meter_elevation asserts the warning fires on a degree-coord meter-elevation raster
• test_planar_no_warn_on_projected_meter_coords asserts no warning on projected meter coords
• Full test_aspect.py passes (200 tests)

Scope

Just the warning. The other aspect findings (dask memory guard, name leak, z_unit error message) are separate PRs.

[brendancol]

PR Review: Warn on degree/meter unit mismatch in planar aspect()

Blockers (must fix before merge)

None.

Suggestions (should fix, not blocking)

One observation, not blocking: the new test helpers _degree_coord_meter_elevation_raster and _projected_meter_raster in test_aspect.py are copies of the same helpers in test_slope.py. The duplication keeps the aspect test file self-contained, which is how the suite is already organized, so it's fine as-is.

Nits (optional improvements)

• The shared warning message in utils.py ends with "before calling slope" even when the call comes from aspect. That wording predates this PR and lives in the shared helper, so leaving it is consistent. A future cleanup could make it backend-agnostic.
• A few existing oracle tests (the cellsize=0.5 variants) now surface this UserWarning because their coordinate spacing looks degree-like to the heuristic. The tests still pass; that's the heuristic doing its job, not a regression.

What looks good

• The fix mirrors slope.py: same helper, same import source, same placement just before get_dataarray_resolution in the planar branch.
• The warning sits in the shared aspect() entry point ahead of backend dispatch, so all four backends (numpy, cupy, dask+numpy, dask+cupy) get it with no kernel changes.
• Tests cover both directions: the warning fires on degree/meter coords and stays silent on projected meter coords. Both pass, and the full aspect suite (200 tests) is green.
• Added a Notes paragraph to the aspect docstring matching slope's.

Checklist

• No algorithm change; warning only
• All backends consistent (warning is pre-dispatch)
• NaN handling unchanged
• Edge cases covered (warn / no-warn both tested)
• Dask chunk boundaries unchanged
• No premature materialization
• Benchmark not needed
• README feature matrix not applicable (no new function)
• Docstrings present and accurate