Only use distance dimensions by default in laplacian for DataArrays

[bhaskargurram-ai]

Description Of Changes

Called on a DataArray with no axes, coordinates or deltas, laplacian adds up the second derivative along every dimension. On a typical model or analysis field with a pressure or time dimension this fails with an error that is hard to read, e.g. for (isobaric, lat, lon) data:

DimensionalityError: Cannot convert from 'kelvin / hectopascal ** 2' ... to 'kelvin / meter ** 2'

Following the proposal in #3105, when f is a DataArray and none of axes, coordinates and deltas are given, laplacian now only uses the dimensions along which positions are distances:

• latitude and longitude, which already go through grid deltas in second_derivative
• dimensions whose coordinate has units of length (projection x/y, or a height vertical coordinate)

Time, pressure and other dimensions are skipped. If no dimension qualifies, it uses all of them as before, so anything that worked before gives the same result (it could only work if all the dimensions had the same units). Explicit axes are used as given. I chose to skip pressure as well as time, since a sum of second derivatives in hPa and in m isn't meaningful. The docstring describes the new default.

Tests added in tests/calc/test_calc_tools.py:

• (isobaric, lat, lon) and (time, isobaric, y, x) DataArrays with no axes now match axes=('lat', 'lon') / axes=('y', 'x'). Both fail on main with the error above.
• a height (km) vertical coordinate is still included, and a DataArray without any distance dimension still uses all of them (both pass before and after, as guards).

tests/calc passes with xarray 2025.10.1. Ruff and flake8 are clean.

Checklist

• Closes laplacian should skip non-length dimensions #3105
• Tests added
• Fully documented

[CLAassistant]

All committers have signed the CLA.