Use of fill when rasterizing categorical data

[gtmaskall]

The default behaviour of make_geocube with categorical data is to append a nodata string to the <varname>_categories variable. This makes sense from a Python indexing perspective inasmuch as the corresponding integer code for this is -1. This seems to override the fill argument. Again this makes sense with the use of -1 and the position of nodata. But if I'm correct in this, the documentation could be tightened up a little to clarify that fill only applies to numerical data.

[snowman2]

Thank you for catching this. A PR with a documentation update is welcome.

[gtmaskall]

First wanted to check my understanding was correct. I'll try to find time.