Index option "unique" should be disallowed for text/geo indexes

The semantics of the {unique: true} option for text/geo indexes are implementation-specific, unintuitive, and not useful (for text indexes, this option enforces that no two documents can have the same term+score pair; for geo indexes, this option enforces that no two documents can have a point with the same geohash value).

This option is currently disallowed for hashed indexes.