CASSANDRA-4511 started adding secondary indexing for collections, but didn't wired map key indexing because this requires generalizing a bit the 2ndary index API to support 2 indexes on the same column. And since that's not entirely related to the initial problem of CASSANDRA-4511, let's tackle this last part here.
I'll note that one other is the syntax. For selection, I propose
(but that assumes we use CONTAINS in
CASSANDRA-4511, if we use IN, we'll need something else)
For declaring indexes I'm less inspired. We could have something like
but maybe someone has a better idea?