Benchmarking runtime of cubulation

Tests are performed on an exclusive compute node of the DKRZ supercomputer Mistral in Hamburg, Germany.

A compute node has the following specs (https://www.dkrz.de/up/systems/mistral/configuration):
  * 2x12-core Intel Xeon E5-2680 v3 (Haswell) @ 2.5GHz,
  * 24 cores (48 logical CPUs),
  * 64 GB main memory.
 
As for the grids we consider limited-area ICON grids that cover a large part of the North Atlantic. They were for example used for the NAWDEX simulations described Senf, F., A. Voigt et al, 2020: Increasing Resolution and Resolving Convection Improve the Simulation of Cloud‐Radiative Effects Over the North Atlantic, JGR Atmospheres. https://agupubs.onlinelibrary.wiley.com/doi/full/10.1029/2020JD032667.

Usage:

benchmark_cubulation.run is a batch job that is submitted to an exclusive compute node via sbatch and that calls benchmark_cubulation.py. The grid resolution, start triangle and search radius for the cubulation are handed over to benchmark_cubulation.py, which reads in the grid file and does the cubulation.
