Support added for the int datatype earth engine image.

[dabhicusp]

The existing code of the Xee is incapable of processing the Earth Engine images which have data of integer datatype. As illustrated in the example below, instead of yielding a "NaN" value, it produces a value that is one less than the ee_mask_value. For instance, if the ee_mask_value is set to -99990, the code will return -99989.

Ex.:
Before :

image = ee.Image('USGS/GMTED2010').select('be75')
proj_4326 = ee.Projection('EPSG:4326', [1, 0, 0, 0, -1, 0]).atScale(10000)
image = image.reproject(proj_4326)
geom = ee.Geometry.Rectangle( [[-179.99, -80], [179.99, 80]], None, geodesic=False,)

ic = ee.ImageCollection([image])
ds = xarray.open_dataset(ic, projection=image.projection(), geometry=geom,
     ee_mask_value=-99999, engine='ee'
)
ds.be75.values

Result:

array([[[     0.,      0.,      0., ..., -99998., -99998., -99998.],
        [     0.,      0.,      0., ..., -99998., -99998., -99998.],
        [     0.,      0.,      0., ..., -99998., -99998., -99998.],
        ...,
        [     0.,      0.,      0., ..., -99998., -99998., -99998.],
        [     0.,      0.,      0., ..., -99998., -99998., -99998.],
        [     0.,      0.,      0., ..., -99998., -99998., -99998.]]])

After updating the code :

ic = ee.ImageCollection([image])
ds = ee_local_class.open_dataset(ic, projection=image.projection(), geometry=geom,
     ee_mask_value=-99999,
)
ds.be75.values

Result:

array([[[ 0.,  0.,  0., ..., nan, nan, nan],
        [ 0.,  0.,  0., ..., nan, nan, nan],
        [ 0.,  0.,  0., ..., nan, nan, nan],
        ...,
        [ 0.,  0.,  0., ..., nan, nan, nan],
        [ 0.,  0.,  0., ..., nan, nan, nan],
        [ 0.,  0.,  0., ..., nan, nan, nan]]])

[naschmitz]

Could you add an integration test case that would've failed before but passes now?