verify_dataset_integrity requires exact equality of image properties when "allclose" is sufficient

[andylitalo]

Running verify_dataset_integrity() checks all input images to make sure they have the exact same spacing, shape, direction, and origin. As a result, if there are negligible (e.g., 1e-10) discrepancies among images, it raises an error when none should be raised.

Suggested fix: make BaseReaderWriter._check_all_same() similar to BaseReaderWriter._check_all_same_array() by replacing this check for exact equality with a check for closeness:

if not np.allclose(i, input_list[0]):

[FabianIsensee]

Thanks for the suggestion. I just changed it to np.allclose in master