Instead of aligning *after* the offset for a buffer view was computed,
we now align each chunk that is written to the binary. This makes sure
that the start of the next chunk as well as the entire resulting buffer
is 4b aligned.