Integration of Worker Runs in extraction

Currently DAN's extraction is using worker versions to identify which transcriptions to get when using the teklia-dan dataset extract with the flag --transcription-worker-version. When a worker version is given there is typically a worker run that it is attached to. For a single transcription-worker-version there could be many transcriptions done by a single version but these worker-versions could be done in different user runs. In this case, I have been told by @yschneider that the program will randomly choose which of the transcriptions to pick from if there are 2 versions. I believe there should be a flag for --user-worker-run to identify what worker run the transcription is from.