fslmaths changes data type to float

if img.nii is uint8, the below code forces float output (not congruent with that the user requested)

fslmaths img.nii -add 0 out.nii -odt input

if using the pass through option (fslmaths in.nii out.nii) the output image is float rather than using the same datatype as the input.

bugs reported by Chris Rorden.