Improve mask performance by selecting only data within shape

[saschahofmann]

Is your feature request related to a problem? Please describe.

I think the performance of the mask function could be greatly improved by only looking at data within the .total_bounds of the provided shape. Personally, I would actually prefer if the returned data would contain only the selected area. E.g. imagine I want to get the country level data for a global dataset. Most of the returned data array will be null. If that behaviour is seen as unexpected for a mask function that maybe providing a clip function that does that would be awesome.

Describe the solution you'd like

Calling something like this before mask = shapes_to_mask(geodataframe, dataarray, **mask_kwargs) (spatial.py L299)

minx, miny, maxx, maxy = geodataframe.total_bounds
dataarray = dataarray.sel({lon_key:slice(minx, maxx), lat_key:slice(miny, maxy))})

should do the trick.

Describe alternatives you've considered

All of this can be easily achieved by calling the code above before calling the mask function, so its no deal breaker. But I would consider it an enhancement. As I said, in the description maybe this would fit better in a clip function than in mask

Additional context

No response

Organisation

Lobelia Earth