[Analytics] Allow opt-in for random_sampler agg

[thomasneirynck]

Describe the feature:

All apps in Analytics (Lens, Maps) should support the random_sample aggregation https://www.elastic.co/guide/en/elasticsearch/reference/master/search-aggregations-random-sampler-aggregation.html

Dashboards should surface a mechanism (a toggle, a slider) for users to trade speed vs accuray.

Describe a specific use case for the feature:

This is important for Dashboards to load fast, regardless of the size of the underlying data. Several apps already make tih trade-off of speed-vs-accuracy (e.g. ML, when doing log-rate analysis, or in the data-visualizer).

[elasticmachine]

Pinging @elastic/kibana-gis (Team:Geo)

[elasticmachine]

Pinging @elastic/kibana-vis-editors @elastic/kibana-vis-editors-external (Team:VisEditors)

[elasticmachine]

Pinging @elastic/kibana-presentation (Team:Presentation)

[thomasneirynck]

Just put this in as placeholder for now. Details forthcoming.

[stratoula]

I am removing the visualizations team assignment as random sampler has been added to Lens!

[thomasneirynck]

We probably should rewrite this a little. The issue with random sampler right now is that has not been designed as a Dashboard level setting. Rather, it is an advanced setting in Lens. This prevents users/authors to apply some global tradeoff of speed against precision.

This should behave more like a global setting. If an author discovers a Dashboard is slow, some kind of toggle/speed setting can be added to "enable sampling".

The main difference is here that random-sampling should not be a property of the chart, but rather a property of the reporting system (ie. the Dashboard).