The handlers for getting original/patched files in the API were

expecting a URL parameter that doesn't exist. It would always attempt to

fetch using the draft filediff resource, instead of intelligently

choosing between that and the published filediff resource.
By using the correct URL parameter name, and using that parameter

correctly, we can fetch the original/patched files correctly.

Manually tested by fetching the original/patched files.
Unit tests pass. We actually don't have unit tests for these resources, though.

Bigger task.