See linked tickets for some ideas, but generally, this ticket captures the work to refactor/improve the structure of the following files:

• cluster_aggregate.cpp 
• cluster_aggregation_planner.cpp
• sharded_agg_helpers.cpp

This ticket is fairly open ended, but some ideas:

• There are quite a few places where we have special cases for change streams, $documents, etc. Perhaps we should consider refactoring these special cases out (i.e.  SERVER-43874), or at least restructure the code in such a way that we don't have a bunch of special case checks scattered throughout the code.
• Similarly, there seem to be separate cases for those aggregates which require a CollectionRoutingInfo and those that don't. We should consider splitting these, as doing so would enable us to acquire the CRI in exactly one spot (as opposed to throughout the sharded agg code path; i.e. SERVER-83658)