Streamline set cache optimization for set operations

In the $project stage of the aggregation framework one set operation ($setIsSubset) has a performance optimization (caching a constant set) which provides tremendous benefits if the set used is big enough.

I think it would be nice to streamline this optimization to all set operations as it is easy to implement and they could benefit from it.

Here is the link to the code implementing this optimization: https://github.com/mongodb/mongo/blob/master/src/mongo/db/pipeline/expression.cpp#L2354