Lazy field list loading

[flash1293]

Related to #134176

Long fields lists can be a problem because currently we always load them up front in their entirety. A lazy loading approach only requesting relevant parts of information and paginating should help with this and speed up the UI even in suboptimal situations (Mapping explosion, Elasticsearch under heavy load, ...)

In general this would require us to treat _field_caps more like a search and less like a way to fetch configuration state: Response might be large, response might be slow, graceful handling of errors keeping the UI functional in a reduced way, ...

I can imagine the following steps:

• Change Elasticsearch to allow pagination on field caps or something comparable and filtering on all properties (type, searchable, …)
• In Kibana, be smart about the fields loaded - do only load those up front that are needed for the vis (but don't cache them, this is causing problems because they could get out of sync)
• For lists (e.g. Lens left hand panel) only load fields page by page while the user is scrolling
• When a field name is searched for, load the fields for this prefix
• When field list filter is applied, do a search for field names

These challenges are similar to those presented by flattened fields - by their very nature there are a large number of them so they have to be loaded on demand.

cc @thomasneirynck

[elasticmachine]

Pinging @elastic/kibana-app-services (Team:AppServicesSv)

[elasticmachine]

Pinging @elastic/kibana-data-discovery (Team:DataDiscovery)

[kertal]

FYI @mattkime, should we ask for the first step, pagination of field caps in elastic search?

[kertal]

FYI @mattkime closing in favor of #152159 some good thought here worth to considering