clean up timeseries access

[matt-long]

ann_avg_utils has unnecessary input parameter:
experiment_longnames

Also, the user pulls things like units out of the module, then puts them back into the model via this API. I would prefer that the routine returns a dictionary of xarray datasets with the units conversion already applied. I am applying the unit conversion post facto:

for var in variables:
    for exp in cesm2_exp:        
        ds = dsets[var][exp]
        
        var_units = ds[var].attrs['units']
        
        var_data_new_units = (
            ds[var].data * pint_units[var_units]
        ).to(final_units[var]).magnitude
        
        ds[var].values = var_data_new_units
        new_units = final_units[var]
        if new_units in pretty_units:
            new_units = pretty_units[new_units]
        ds[var].attrs['units'] = new_units

This routine doesn't seem to actually be computing timeseries, but rather constructing cache file names. I would prefer that the computation happens, if necessary, here.

Must we do any concatenation inside this function? It seems like mission creep.

[matt-long]

Presuming it's not replaced by a broader revision, I think get_ann_means_and_units should be renamed to something like get_timeseries and possibly accept a freq argument.