Metadata lookup

Many times people just download a PDF and want to automatically retrieve the metadata (say, on Zotero import). It would be cool to provide a service mapping `md5sum(pdf)` → canonical URL. We could implement this even in our static context just by creating a directory of symlinks from `https://aclweb.org/anthology/hash/X` → `https://aclweb.org/anthology/Y`. This could be computed separately in the same way we currently build paper thumbnails.