When processing a new file conversion and text extraction is skipped
if detected to be already done. This prevents running expensive tasks
again after restarting/retrying. When explicitely reprocessing a file,
these tasks should run again and replace the existing results.