- Use VLC instance to parse metadata from the media files.
- Switched to MD5 for hashing the files.
- Switched to using the file location as the proxy for if we have seen
the file rather than hash. Hashing every file was slow.
- Moved away from using jsonutils {to, from}Json methods for persistence
of the libraryt DB. For simple objects it works well, but serialized
the internal implementation of tables, etc. Now using a hybrid. Still
use jsonutils for a number of the models, but have custom code to
serialize TableRefs and other data structures.