Use elastic search directly for timeseries column names

[HansKallekleiv]

For discussion.

• Use elastic search directly instead of the sumo explorer
• Get the column names from the first valid realization instead of the collection.

Reduces the time spent from 15-20s to 0.5s. This is probably due to how how the explorer/sumo wrapper uses batching when handling large document collections from elastic search. Should be investigated.

[anders-kiaer]

Superseded by another PR