Now that implemented image scaling, the processed images currently don't contain metadata. We need to put appropriate metadata in the output files. To large extent this will be the reverse process of GUISE-148.
We should probably be more selective about which metadata is used and what it is translated to. Only using XMP may not allow Windows Explorer to see the metadata; but Exif for Windows Explorer doesn't consistently handle UTF-8 on all consumers. Maybe we don't care about other consumers, though—maybe we just care more about archiving. But if the image is downloaded from the web, it would be useful for consumers to easily see (e.g. in Windows Explorer) if there is metadata—especially rights metadata.
One solution might be to convert any string to ASCII for saving to Exif. We could drop accents and convert e.g. "©" to "(C)". But dropping any accents is not so good an idea; maybe we should try to go with UTF-8 anyway. (See discussion in GUISE-148.) I'm not sure offhand which libraries support writing image metadata, though—I don't think the metadata-extractor library handles this.