STYLE: Encapsulate NiftiImage qto_xyz mat44 data as Matrix<float,4,4>

[N-Dekker]

Using itk::Matrix<float, 4, 4> (instead of std::vector<float>) to store the "qto_xyz" data from m_NiftiImage, as this Matrix template instantiation more closely resembles the original niftilib mat44 data structure.

Follow-up to pull request #3242 commit 9eb4c12 "ENH: bug #3241: Add qfac and qto_xyz to itkNiftiImageIO metadata" by Sean McBride (@seanm), which was merged on March 17, 2022.

[seanm]

OK, give me a chance to try this in my app, want to make sure it actually works for me...

[N-Dekker]

OK, give me a chance to try this in my app, want to make sure it actually works for me...

Thanks @seanm I'm not in a hurry 😃

If you want be sure than it does not get merged accidentally, you may do Review changes -> Request changes, and then switch to Review changes -> Approve once you're sure it works for you.

BTW, instead of using Matrix<float,4,4>, we might now also store the data of "qto_xyz" as a raw C-style array float[4][4], if you would like that better. But I think using Matrix<float,4,4> is most convenient in this case, as itk::Matrix<T,Rows,Columns> is a regular, copyable C++ class type, supporting the most common matrix operations, and interaction with vnl_matrix

[seanm]

OK, I've tested, this is good for me.

(Calling it "STYLE" seems a little light though, it is a behaviour and public API change, even though the old way didn't ship in a release.)