Pipeline aggregations are weird!

• They are validated on the data node rather than the coordinating node. (Move pipeline agg validation to coordinating node #53669 + More pipeline aggregation cleanup #54298)
• They are built on the data node and shipped back to the coordinating node (Save a little space in agg tree #53730 + Fix pipeline agg serialization for ccs #54282 + Begin to drop pipeline aggs from the result tree #54311 + More pipeline aggregation cleanup #54298)
• Because of that they implement NamedWriteable, equals, and hashcode. None of that is really important if we stop doing the above. (Deprecate serializing PipelineAggregators #54926 + Remove serialization from pipeline aggregator #55026)

There are a few other things we could probably clean up here but this'd be a huge step for making pipeline aggregations simpler to implement. And, bonus, it'd save some space in the aggregation result tree.