Feature ask: sibling version of bucket script aggregation

[monfera]

Describe the feature:

Bucket script aggregation is a parent pipeline aggregation and it provides flexibility for when the premade pipeline aggregations won't cover some need.

For symmetry, and to meet the data requirements of a broad range of of visualizations, it'd be useful to make it alternatively work as a sibling aggregation.

Here are a couple of visualizations (I'm not suggesting that these can't be done with other, more targeted sibling aggregations or with distinct, N-ary parent aggregations, so they just illustrate the structure of the data need):

Heatmap with marginal histograms: the histograms are just aggregations (along two different fields) of the heatmap part

Scatterplot with boxplot: the boxplot height/position etc. can be computed as statistics over the scatterplot points (which themselves may be aggregates)

Basically, any kind of multilayer charts; eg. here the faint points are the most granular aggregates, and everything else including the bars are aggregations over them, this is a random prototype for a multilayer visualization (I'm not sure if the exponentially weighted moving average line and the Bollinger band around it is feasible, but the others likely are)

Entered on encouragement by @polyfractal

[elasticmachine]

Pinging @elastic/es-analytics-geo (:Analytics/Aggregations)

[wchaparro]

Added to pipeline refactoring Meta: https://github.com/elastic/elasticsearch/issues/82808

[elasticsearchmachine]

Pinging @elastic/es-analytical-engine (Team:Analytics)