crypto_core_ed25519_from_uniform exposes the *direct* map

[LoupVaillant]

From the manual

This function directly exposes the inverse Elligator 2 map

The original paper however says otherwise:

• representative to curve is the map.
• curve to representative is the inverse map.

Is this a typo, or is there a consensus about the paper being "wrong"?

(Edit: it wouldn't be the first time usage diverges from the original paper. Montgomery curve points for instance now tend to be called (u, v), while the original papers used (x, y).)