Bugfix CV: mask out nan nodata

[davemfish]

We were not properly masking out nan values from the bathymetry raster.

I updated an isclose check and updated a test to confirm that nan values can be properly masked like any other nodata value. Other raster ops in this model were already doing the proper checks via pygeoprocessing.array_equals_nodata.

Fixes #1528

Checklist

• Updated HISTORY.rst and link to any relevant issue (if these changes are user-facing)
• Updated the user's guide (if needed)
• Tested the Workbench UI (if relevant)

[phargogh]

Makes sense to me, thanks @davemfish !