as far as i know, this can’t be done with the standard CSV export currently. you’d probably have to use the API to get a set of observations with both numeric ID and UUID, but it’s not very efficient if you have more than 10,000 records in your set.
it looks like the DWCA export files that go to GBIF, ALA, etc. can include UUID, but they either don’t or aren’t processed by the receiving system(s).
the only other thing i can think of is, if iNat development staff don’t get around to it, and you have the skills to do so, you could update the iNat code to get the UUID into the CSV export.