LoadImage/SaveImage consistency with itk reader

[wyli]

Describe the bug
Saving a Dicom image into Nifti, and load the nifti, we should get the same data representation.

To Reproduce

import itk

from monai.transforms import Compose, EnsureChannelFirstD, LoadImageD, SaveImageD

img_dir = "tests/testing_data/CT_DICOM"
data_dict = {"img": img_dir}
keys = data_dict.keys()
xforms = Compose(
    [
        LoadImageD(keys, reader="itkreader", pixel_type=itk.UC),
        EnsureChannelFirstD(keys),
    ]
)
img_dict = xforms(data_dict)  # load dicom with itk

save_xform = SaveImageD(keys, meta_keys="img_meta_dict", output_dir=".")
save_xform(img_dict)  # save to nifti

new_xforms = Compose(
    [
        LoadImageD(keys, reader="itkreader"),
        EnsureChannelFirstD(keys),
    ]
)
out = new_xforms({"img": "CT_DICOM/CT_DICOM_trans.nii.gz"})  # load nifti with itk
print(img_dict["img"].shape)
print(out["img"].shape)
print(img_dict["img_meta_dict"]["affine"])
print(out["img_meta_dict"]["affine"])
print(img_dict["img_meta_dict"]["spatial_shape"])
print(out["img_meta_dict"]["spatial_shape"])

inconsistent output:

(1, 4, 16, 16)
(1, 16, 16, 4)
[[   0.488281      0.            0.         -125.        ]
 [   0.            0.488281      0.         -128.100006  ]
 [   0.            0.           68.33333333  -99.480003  ]
 [   0.            0.            0.            1.        ]]
[[ -0.48828101   0.           0.         125.        ]
 [  0.          -0.48828101   0.         128.1000061 ]
 [  0.           0.          68.33333588 -99.48000336]
 [  0.           0.           0.           1.        ]]
[ 4 16 16]
[16 16  4]

this is an integration issue of nibabel writer <=> itk reader