Allow to optionally perform extraction on full page raster images in PDFs

This is off by default as it requires quite heavy image processing, but
if an application is certain the input PDF contains something that can
be extracted this can be enabled.

Makes importing Danish COVID certificates work.