Another possibility is one of their programmers thought "It would be good if there was more encrypted e-mail going around in general, I wonder if I can get it into facebook somehow" and coded this feature in their free time. Then convinced his managers to integrate it with that argument plus "and it's already coded we just need to merge it in"
But I could certainly see some benefits both for FB and for world at large from this. One of the big problems with PGP is how to bootstrap web of trust. "Does this key really belong to this particular person?" But what if the otherwise loathed real name policy could be turned to service this particular need? Prominently visible personalities can attach their PGP keys to their pages and make the first association harder to forge.
Secondly, I have little doubt that the keyservers are monitored. An increase of searches and/or downloads to known activist lawyers' or journalists' keys could have relation to uncomfortable whistles being blown in near future. But what if FB made the keys they have signed available via their own keyserver, and made that reachable over Tor? Downloading a high-profile PGP key is likely to be a fairly big red flag.
And lastly, there may be some positive effects further down the line. I've been using PGP (and later GPG) since 2.3i became available and I know just how horrid the usability is. If FB can iterate over UI and UX issues, then others can learn from those efforts, and eventually we might have something that even a regular person could at least learn to use.
And of course - adding more encrypted noise to global email flow is not a bad thing at all.
I have no doubt that FB sees many non-altruistic avenues if this service catches wind. Wonder is there is anything to relationship graphs with some extremely strong edges...