Since we're just importing a single vector, we don't have to turn it
into an ITER_IOVEC. Instead turn it into an ITER_UBUF, which is cheaper
to iterate.
Signed-off-by: Jens Axboe <axboe@kernel.dk>