Fix the ambiguity about percentile scaling in image saver

[dkazanc]

Currently there is an ambiguity between the calculation of the percentile and the percentage when the glob_stats argument is provided OR not. Two different scaling approaches are somewhat unified.

It makes things even more confusing because of the dependency on the glob_stats argument and data rescaler function in httomolibgpu package. This rescaling function might be used before save_to_images and then there is no need to provide glob_stats to image saver to rescale the rescaled data again! However, when the glob_stats argument is NOT provided, we then use a time consuming percentile rescaling without any need for it.

Also if you rescaled once with rescale_to_int and then provide glob_stats to save_to_images, you will rescale your data again, incorrectly, I believe.

[dkazanc]

HTTomo solution: One possible solution is to make rescale_to_int to be a part of image_saver, however, then it must become a CPU method. Another solution is to make rescale_to_int insertable before save_to_images in HTTomo automatically based on the parameters given in the save_to_images. Slightly fiddly but the obvious thing here is that those methods need to know about the existence of each other so to avoid duplications of operations. Any other ideas?

[dkazanc]

17