Add magic number classification in the FileMetadata cataloger/indexer

Today we report the file type as described from the tar header for each file in the image. We do not classify beyond a "regular file" what is may be. It would be interesting to also include MIME type classification (or similar) based on the magic number for scanned payloads. Most magic numbers can be found well within the first sector of reading, so the entire contents do not need to be scanned.

Start at the FileMetadata cataloger/indexer and review how file types are currently taken from tar headers. Define the magic-number and MIME classification scope, including the bounded read needed for scanned payloads. Done means regular files can receive a useful classification without scanning their entire contents.

Written by the indexing model from the issue text.