Get unique event id for empty df

I'm skimming based on event IDs for my chunks to get a BCS validation tuple. This requires accessing unique_event_id which might not be set if I happen to come across an empty chunk (rare, but happens). Could of course handle this in my code, but I think as a general rule functions that set columns should set them even for empty dataframes to avoid this trouble.

It's really annoying that df.loc[:, col] doesn't work with empty frames.