Base chunk size on number of dask processes

[martin-springer]

Is your feature request related to a problem? Please describe.
To improve computational efficiency, the chunk size should be somewhat correlated with the number of dask processes.

Describe the solution you'd like
We need to re-chunk some of the dask arrays during an analysis. Currently, this is done based on arbitrary chunk sizes.

Describe alternatives you've considered
Obtain the number of processes from dask cluster and design chunk size based on that.