can we generalize speckle` _process` so that bincount can be used on floats?

See the issue below, as they are related, but this is to address a specific question of
np.histogram versus np.bincount raised in
https://github.com/NSLS-II/pyCHX/issues/52

np.histogram is used for image based data because image normalization process creates floats (this is the case for csxtools get_fccd_images()). Some discussion of the various options would be good. Please add other options if I have missed one. We can consider multiple approaches too, but probably need to prioritize

1. have 2 _process subfunctions: one for integer data (like Eiger detector) and one for float data
2. convert floats to integers in a way that we agree.
3. require input (whether images (aka full np.array) or compressed data) must be integers

@danielballan and others?

Start by reading linked issue #52 and locating the _process subfunctions that use np.histogram or np.bincount. Compare the three proposed approaches for integer and float image data, then confirm the chosen behavior and scope with maintainers. Done means the agreed approach is implemented and its behavior for image and compressed data is covered.