Reduce diff storage by hashing diff uploads

Diff files that already exist in the database will no longer be double-stored. Diffs are now hashed on upload and correspond to a new hash->binary table.
Existing table data is not hashed and will remain for backwards compatibility. Evolutions have been made to create the new table and rename existing fields, so that model-logic can override fields. Test data has also been modified for new field name compatibility.

Ran fill-database and all data was hashed accordingly. Manually posted reviews and posted parent diffs. Created a unit test to check if hashes match.