FIX: gracfully handle missing seaborn dependency

[scott-huberty]

xref: #9561 (comment)

Fixes a regression I introduced in #9561

MWE:

On current main, this code will fail if seaborn is not installed even if matplotlib is installed:

conda create -n xr-test python=3.10 xarray pooch scipy matplotlib

import xarray as xr
air = xr.tutorial.open_dataset("air_temperature")
air.air.isel(time=0).plot.contourf(colors="k")

As mentioned in #9561 (comment) , it seems that this wasn't caught by a CI because there isnt' a CI that has matplotlib but not seaborn installed.

EDIT: AFAICT - The behavior before #9561 was that if you pass a single color to the colors parameter like in the MWE, xarray will try to use Seaborn for this but fallback to matplotlib if necessary (e.g. if Seaborn is not installed).

[max-sixty]

Thanks a lot for checking & following up!

[scott-huberty]

Thanks a lot for checking & following up!

No problem, figured I should fix what I broke! 😅

[mathause]

Thanks!