Dump calo clusters

Adding jet constituents to GN2 has some issues (starting with #124 (closed)). We could try going a level lower and directly adding calo clusters. We loose the energy subtraction that goes into the jet constituent building, but maybe we can do better anyway with a fully learned approach. Calo clusters might also be easier to access in athena during inference than following all the links through the jet constituents, and taking care of the track selection overlap issues.

Calo cluters inherit from IParticle so I think they shouldn't be too hard to dump using our existing writers.

cc @jabarr @ioleksiy