My impression is that the .mht files are MIME messages, of type multipart/related. A MIME parsing module, like MIME::Parser should work to split it apart. I don't know what content encoding they used; the binary files may not be encoded.

The two option I see for storing the files are put them in a direcotry, or put them in archive of another format, probably .zip would be best. The advantage of a directory is that they should viewable as they are. The advantage of zip archive is that you get one file that normal tools can use, but can't be directly viewed by browsers.