Update binary references to use extracted copies, rather than originals
Things like embedded images, media, and math are all stored in the .docx directory. For the HTML extraction, these files should be copied over to the same directory as the HTML files. That way, they're easily accessible, and the HTML doesn't require the input .docx file to stay where it originally was. However, XSLT doesn't allow for file system manipulation by itself. That task will fall to INK, which is slated to be rebuilt in JavaScript (rather than RoR). Once that is complete, XSweet should be updated to reference copies of the binaries in the output directory, rather than directly referencing the binaries of the original .docx file.