Group the results of save_datasets by output file, when multiple files
are being written and each file has multiple RIODataset or RIOTag
objects.  This is helpful when we want a wrapper per file and therefore
a single ``da.store`` call for each file.

Add an implementation to split sources/targets by file.  This helps if
we want to have a wrapper function to execute immediately when each file
is completed when time for copmutation comes.

Adapt call to zip for compatibility with older versions of Python

Add an example on how to use the utility function
group_results_by_output_file.  Also add a warning that for large
calculations, this appears to cause a slowdown.