Trust debian/copyright

Go through the scanning results, trying to check each UNKNOWN entry against debian/copyright. Instead of (naïvely) pasting the found copyright and license info directly into that paragraph, store that info aside, delete the original paragraphs and try to collate the newly generated paragraphs so that files with identical copyright/license are all grouped into one paragraph for each such combination, rather than having one file in their own paragraph. This reduces the size of the generated file from dozens of megabytes to hundreds of kilobytes for some packages.