[Lens] Do not use "Top values" agg for number fields in tables

[flash1293]

Currently, when dropping a number field in the empty dimension trigger of the data table, it will add a "Top values" column by that field. In most cases this is not the right thing to do, as number fields are commonly shown as a metric.

In this scenario, an average operation for the dropped number field should be added.

[elasticmachine]

Pinging @elastic/kibana-app (Team:KibanaApp)

[wylieconlon]

I think this is possible within the current suggestion system if we make this change:

1. The datasource would generate more than one suggestion which uses the number field, it would generate both a metric suggestion and a bucket suggestion. We might need a new suggestionType for this, but probably not.
2. Visualizations like the XY chart should update their suggestion logic to be smarter about which fields are bucketed or not, so it can continue ranking the metric suggestion higher
3. The datatable visualization could rank buckets higher

[flash1293]

I'm just realizing - As this is happening when you are dragging the field into the dimension trigger on the right side, is the suggestion system even used for that operation? Looks like it simply calls

kibana/x-pack/plugins/lens/public/indexpattern_datasource/dimension_panel/dimension_panel.tsx

Line 109 in bf04235

export function onDrop(

[wylieconlon]

I think you're right that there is inconsistency with the onDrop function and the overall suggestions- this is a piece of tech debt that has caused bugs before, for example when the onDrop was not selecting the highest-ranked function.

[flash1293]

Maybe we have to refactor this to use the suggestion system in that case as well to solve it in a nice way.

As a stopgap measure we could prefer number field metrics to number field buckets if both are possible by the given operation filter. It wouldn't change any apis but it would be pretty hacky within the datasource onDrop logic because the type of the underlying field shouldn't matter to which operation is picked.

Instead of picking the first one here:

kibana/x-pack/plugins/lens/public/indexpattern_datasource/dimension_panel/dimension_panel.tsx

Line 152 in bf04235

op: operationsForNewField[0],

check whether the field has a number type and if that's the case, check whether there's a possible metric operation as well.

[flash1293]

This is no longer happening - for metrics it will default to median (which makes sense), for buckets it will default to histogram