Implement non-FHIR type in population criteria

[mdnazmulkarim]

Currently, the population criteria must return a type consistent with the population basis, which is only allowed to be a FHIR type.
We can consider an option to return a non-FHIR type such as Tuple.

Also there needs to be another issue on the evaluator to correct the issue with non-patient-based measures not computing intersection of results