Functionalities required to finalize documentation and generate API

[rhugonnet]

It looks like we are missing these functions:

• Vector.save()
• Vector.show()
• Vector.__str__()
• Vector.__repr__()

And these properties are not defined as such for the API:

• Vector.ds,
• All of additional SatelliteImage attributes,
• Raster.crs and Raster.transform.

Also, looking at our geospatial tools, looks like we could add:

• Vector.simplify() to simplify/degrade resolution of a Vector.
• Make indexing [] possible with boolean ndarrays.
• Allow to pass a Mask in Raster.set_mask(),
• Make "in_value" and "target_value" attribute name consistent between polygonize and proximity.
• Make equal_raster and georeferenced_grid_equal naming more consistent (start with equal for both?)
• Automatically load data when calling arithmetic, logical or NumPy operation? (remove some load_data=True in the docs due to this)
• Make show() plot on top of same ax by default.
• Add clip option to Vector.crop().
• A subsample wrapper function directly in the Raster class (allow to return coordinates).
• Ensure support of Ufuncs through np.ufunc.reduce and other similar operations,
• Rename cropGeom into crop_geom for consistency with other argument syntax,
• Streamline main description of Raster and Vector methods and attributes,
• Add tests for Mask logical operations and overloaded methods,
• Change nodata from list to tuple (needs to be immutable as a property),
• Add tests for untested functions (Raster.shift(), Raster.to_xarray(), etc),
• Improve general test coverage, in particular checking warnings raised directly by GeoUtils.
• Added as_array option to create_mask, as it nows returns a Mask by default. The shape is also always 2D to avoid the squeeze previously used everywhere.
• Fixed loading multi-band raster explicitly after instantiation with load(), failed because of how out_shape was defined.

Bug found:

• Maximum recursion depth calling np.ma.mean on a Raster.
• In doc, array interface: isclose is returning a masked_array instead of Mask?
• Cast from Mask to Raster does not work in Mask.proximity, need to modify Raster.from_array?
• Fixed coords(grid=True): add more tests

Changed (need to add tests in the test suite):

• Defined Raster.__repr__, modified __str__ to be consistent with NumPy,
• Set load_data=False as default,
• Loading data by default if stats=True is passed to Raster.info(),
• Cast array to float automatically in get_nanarray(), otherwise was raising error,
• Added overload for proximity and polygonize for Mask,
• Changed default parameter of polygonize to "all".