Rescale to int cpu/gpu agnostic

[dkazanc]

CPU-GPU rescale_to_int function. All tests modified to compare the results of CPU with GPU results.

I discovered that the GPU function accepts only float32 array as an input yet the CPU one can accept any data type (because of the cast to float64). The GPU implementation is significantly more memory efficient because of the elementwise loop and casting scalars instead of arrays compared to Python one-liners.

So far I don't see the reason to satisfy other data type inputs, however, in future we might work with uint16 then this code must be adapted accordingly.

I'll modify image saver in here accordingly, we need to remove redundant parameters.

[yousefmoazzam]

Apologies for the broken IRIS CI job, it looks like maybe IRIS isn't happy, the conda env failed to be created... https://github.com/DiamondLightSource/httomolibgpu/actions/runs/11052312862/job/30704122840?pr=157

[dkazanc]

Detailed docs page added
https://diamondlightsource.github.io/httomolibgpu/reference/methods_list/rescale_to_int.html

Seems like IRIS is down, I'm merging this PR...