Previously, the data segments were being aligned twice.
This caused us to overalign the segment and therefore allocate a much larger
size for each segment than was required. This fix ensures we align and set the size
just once, ensuring semantically correct binaries as well as smaller binaries.