[Lens] Allow aggregation over time series for all functions for mode time_series data views

[flash1293]

When working with time_series data, it's very common to do a metric calculation per series (e.g. the average of a value or the rate at which it's growing), then rolling up / aggregating the various series into a single unified number. Most often this aggregation is summing up the various values. For example "Get the current memory usage for all processes and report the total number" (Last value of the memory usage field per series, then adding up all of these values).

To allow this, a new "Aggregate individual time series" option should be added to the metric configuration in case the field is only used in time_series indices

This will translate into running the metric agg within a time_series aggregation, then using a sibling pipeline aggregation to aggregate the results per time series:

date_histogram {
  time_series {
    my_metric: { < picked function (e.g. average) > }
  }
  < picked aggregate (e.g. sum_bucket / avg_bucket) > > time_series > my_metric {}
}

This option should not be available if the current data view is mixed time_series and non time_series data for the selected field.

[elasticmachine]

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