Batch runs appear to mix together active measures in scenarios that have the same prep settings but apply to different measure subsets

[jtlangevin]

When executing a batch run across two cases that draw from the same measure set but use different (but overlapping) subsets of the larger set, the batch runs of ecm_prep.py will develop a shared set of prepared measure data for run.py to draw from in each case, and will avoid preparing the same measure twice across cases that share it, as intended.

However, then each case's execution of run.py pulls in the full set of prepared measures, rather than the appropriate subset of measures that the case applies to.