Numpy representation of Rasters causes issue with arithmetic overloading

[adehecq]

This is a bug on a yet non-existing code, but related to on-going work on issue #73.
The idea is to enable rout = r1 + r2 with r1 being a Raster and r2 being either a Raster, a numpy array, or a number. This should return a Raster.
This works great in most cases, except in the situation when r1 is a numpy array and r2 a Raster. In that case, numpy's __add__ method is used instead, which returns a numpy array. This is counter-intuitive in my opinion and should be avoided.
I believe the reason behind it is the numpy's interface, related to PR #210. When issue #73 is solved, one needs to check how to have both features working together.

[erikmannerfelt]

Is this still an issue?

Also perhaps this should be done in conjunction with #261

[rhugonnet]

This actually works now with #302 (the numpy interface is not taking over anymore):

test = np.ones(r.shape) + r
type(test)
Out[15]: geoutils.georaster.raster.Raster
test = r + np.ones(r.data.shape)
type(test)
Out[17]: geoutils.georaster.raster.Raster

However, as you can notice, I have to change the shape to .data.shape on one side, and not another, otherwise:

ValueError: Both rasters must have the same shape.

I'll leave this issue to be closed by #302 and open a new one on this.