use zarr as base file format instead of HDF5

[glyg]

HDF5 has issues - it is not very parallel friendly, there is only one implementation of the interface
Also, we don't store the metadata very nicely now (in a json file, not always there, etc.) it would be a good occasion to rethink our I/O workflow.

Along those lines, it would be good to have a more formal simulation workflow - intialization, settings, run the sim, etc.

[glyg]

Following discussions with regard to the use of zarr for biology images, it seems that maybe using NetCDF would be a good idea, as it is a proven abstraction layer.