Parallelize generation on maps

On large datasets the generation of maps takes an extremely long time.

Ran generation of avg maps on the Snorre field took 8 hours to produce close to 500 maps.

Introducing async and possibly multiprocessing should cut that time significantly

The issue names no files, tests, or entry points. Start by locating the map-generation entry point and measuring the large-dataset path described for the Snorre field; determine whether async or multiprocessing fits the workload. Done means map generation is parallelized and the reported long runtime is materially reduced without changing the generated maps.

Written by the indexing model from the issue text.