Rather than try and change the way your CSV is structured, maybe look into rewriting the vanilla import parser so that you can read your CSVs? That way you’re stacking the work with your own strengths (understanding your CSV format) rather than trying to dabble in the unknown (spelunking in Discourse’s DB schema).
If you’re intent on importing directly to postgres, I don’t have a good working knowledge of the schema to answer your question directly, sorry. I’m just throwing some alternative suggestions your way to solve your problem.