Hi Leonidas, I've been meaning to respond to this thread for a few days, always thinking I'd have a little more time to prepare... At any rate, no, the RDM in v4/5 has never imported CSV files (it more or less did in v3). However, a long time ago I made a conversion utility for a project I was on. You can find it here, somewhat buried within a v4/5 Arches package:
https://github.com/legiongis/fpan-data/tree/master/utils. The top half of the README is relevant to you, and the scripts and sample data are inside of the "reference_data" directory. Feel free to copy those files out to your own repo if you want. Also note that I am in the process of making some changes related to the business_data side of the utils directory, so some updates may change the way that README looks in the future.
If you can get the scripts working, they take a set of CSV files and convert them into a single thesaurus, and a set of collections that mimic the structure each CSV. So, if you have a CSV called Styles, a new Top Concept called "Styles" will be created and then the contents of the CSV will be added as individual member concepts within that top concept. Then a new Collection is made by the same name and the concepts are added to it as well. Note that this is a naive structure, and doesn't support nested concepts, which I know many thesauri rely on (mine for that project didn't so I was ok with that limitation).
Hope that is somewhat helpful,
Adam