Lower Memory Consumption while Evaluating Stratifiers

[alexanderkiel]

Currently stratifiers are evaluated after all populations of a group of a MeasureReport are evaluated. The population will consist of a list of population and subject handles in that case. That list can get very big on large populations. In a use case were 52 million Conditions should be stratified by their codes, a heap size of 128 GiB was necessary.

Instead of creating a list of all subjects of a population, the stratifiers should be calculated on-the-fly and the result should be the finished MeasureReport group.