Fix device mismatch errors in torchstain color space conversion functions

[SamirMoustafa]

The rgb2lab() and lab2rgb() functions in torchstain/torch/utils/ were failing with RuntimeError: Expected all tensors to be on the same device when processing CUDA tensors.

This occurred because:
Constant tensors (_rgb2xyz, _white, _xyz2rgb) were created on CPU by default

Input tensors were on CUDA devices
PyTorch operations between tensors on different devices are not allowed

Solution
Modified both functions to automatically detect the input tensor's device and move constant tensors to the same device:

1. rgb2lab.py: Added device detection and moved _rgb2xyz and _white tensors to input device
2. lab2rgb.py: Added device detection and moved _white and _xyz2rgb tensors to input device

[carloalbertobarbano]

Thank you for your contribution @SamirMoustafa. I will merge the PR to the development branch.