Good question.
I suppose Ferennil, Ferennis would be fine too, perhaps even better, but I do think Ferendir, Ferendis are at the very least possible.
Suffixes coming from ✶ⁿd- seem to have somewhat inconsistent behavious in Sindarin. We have examples of both -nn- and -nd- (and I am ignoring cases of -ndnd- shortening to -nd- as in Brandir, those seem consistent):
-
randir, condir, Borondir, (Telchandir?) on the one hand; and
-
Edennil, Gelennil, Gaerennil on the other.
And then there are also Aerendil/Aerennel, both given as phonetic adaptations of Earendil (the proper Sindarin form of which is given as Gaerdil or the aforementioned Gaerennil).
One might speculate that the retention of 'nd' after a stem ending in -n might be a feature of later Sindarin, whereas the more historically "proper" form would have 'nn'.
But at the very least I can confidently say that if Mithrandir is a licit form, then so is Ferendir. And, in the case of Ferendir, being identical to its Quenya cognate is a major bonus in my book :D