In some of the OptaPlanner examples, we have Constraint Streams like this:

                .groupBy(ProcessAssignment::getService, count())
                .groupBy(Service::getTag, max((service, count) -> count))

This will:

• Take all ProcessAssigment instances.
• Count all such instances which run on the same Service.
• Out of all the pairs of (service, count) find the maximum count where the services share the same Tag.

This is an example of groupBy composition. Currently, we implement it with the old, inefficient groupBy. That implementation looks something like this:

	when
		$tuples: List<Tuple<Service, Integer>> from accumulate(
			$p: ProcessAssignment(),
			...
			groupBy(ProcessAssignment::getService, count())
		)
		$tuples2: List<Tuple<String, Integer>> from accumulate(
			$tuple: Tuple<Service, Integer> from $tuples,
			...
			groupBy(Service::getTag, max($tuple._2))
		)
                $result: Tuple<String, Integer> from $tuples2

	then
		System.out.println($result);
	end

We'd like the new groupBy construct to be capable of composition instead.