CAS multiplier phase support #21
Description
Currently, the CAS management strategy is fixed - each phase will create a single fresh CAS with a copy from the preceding phase. However, this does not seem to be practical for some purposes, for example QA pipelines.
In a QA pipeline, it would make sense (akin e.g. to the publish DeepQA approach) to create a separate CAS for each search result, then run a pipeline within that CAS that will create a separate CAS for each candidate answer and run a pipeline for that, etc., and eventually a consumer that will juggle all the generated CASes. However, this does not appear to be easily possible with the current setup.
Perhaps this is already possible within the framework - in that case, I'll appreciate some directions.