Normalise Experimenters, Groups, Details

[will-moore]

It would be great if omero-marshal could ease the burden of normalising owners, groups and possibly even details too, to reduce the size of the python dict / json object that is needed.

This would need an encode() to be able to take a list of IObjects and return either a tuple of 3 dicts: (obj, owners, groups) or a single dict that included all three.

My current simplistic approach in ome/openmicroscopy#4708 for normalising encoded dicts doesn't handle nested objects:

def normalize_objects(objects):
    experimenters = {}
    groups = {}
    objs = []
    for o in objects:
        exp = o['omero:details']['owner']
        experimenters[exp['@id']] = exp
        o['omero:details']['owner'] = {'@id': exp['@id']}
        grp = o['omero:details']['group']
        groups[grp['@id']] = grp
        o['omero:details']['group'] = {'@id': grp['@id']}
        objs.append(o)
    experimenters = experimenters.values()
    groups = groups.values()
    return objs, {'experimenters': experimenters, 'experimenterGroups': groups}