Metaqueries to list field values take slow path when the only predicate is _measurement

[williamhbaker]

In #19856, an alternate path was implemented for retrieving values from the store to satisfy queries for getting tag values for _field, where there is a predicate on anything except _measurement (see associated issue: #19794). This alternate path involves scanning through all series in the shards within the specified time range, and is considerably slower than the standard path.

A common metaquery used by the UI is to retrieve all fields for a certain _measurement. It might look something like this:

from(bucket: "bucket")
  |> range(start: -30d)
  |> filter(fn: (r) => (r["_measurement"] == "measurement_name"))
  |> keep(columns: ["_field"])
  |> group()
  |> distinct(column: "_field")
  |> limit(n: 200)
  |> sort()

This query has a predicate only on _measurement, and should not have to take the alternate path. However, the current implementation does cause this query to take the alternate path and thus potentially take longer to execute than it would need to.