Gantry tilt sform

[neurolabusc]

Paul Wright noted that CT scans with gantry tilt look distorted when the shear is reported in the SForm, but looks correct in the resliced _tilt corrected image. This extends issue 253.

This problem is easy to spot in the dcm_qa_ct phantom images, where the spherical ball appears ellipsoid and does not match the resliced image (see left panel below). The latest commit to the developmental branch attempts to resolve this by adjusting the formula for the sform (right panel). Note that SPM adjusts the bounding box depending on the SForm, but in reality the _tilt images (lower row in image below) are identical.