Add flag to convert PDF images to grayscale during loading already

This avoids having to do it during binarization for barcode decoding, at
which points its a bit more expensive, and it reduces the amount of memory
we need in the process.