Bug fix in Pytorch nnUNet preprocessor.py

[aecelaya]

In PyTorch/Segmentation/nnUNet/data_preprocessing/preprocessor.py, line 220, you use the np.argmin function to find the low resolution axis for the target spacing variable. However, the low resolution axis is the position of the largest component of the target spacing. Hence, the np.argmax function is needed to find the low resolution axis.