Decide of the naming of strict or relaxed equality (__eq__ or raster_equal?)

[rhugonnet]

See #313 (comment)
Right now __eq__ checks that data.data and data.mask (and other attributes) are equal, and np.ma.allequal can be used to compare .data with masked values.
Several solutions we could implement:

• We could add np.array_equal to __array_function__ to be castable on a Raster;
• We could add a gu.raster_equal function that accounts potentially for nodata with an argument;
• For equality with __eq__, what do we do? Should we mirror behaviour from NumPy or not for this? Actually NumPy broadcasts to element-wise equality with __eq__.

Example copied from @adehecq on Slack:

import numpy as np
data = np.ma.masked_array(np.arange(9).reshape((3,3)))
data[0, 0] = np.ma.masked
data2 = data.copy()
data2.data[0, 0] = 10
data == data2. # returns True for all unmasked values, masked otherwise
np.all(data == data2) # True
np.ma.allequal(data, data2) # True

[rhugonnet]

Solved by #351