storeByGroup improves collector performance immensely in IO constrained systems.
In order to take full advantage, we could use a couple more features.
1. A convertor, to import existing single datasource RRDs to storeByGroup style
multiple datasource RRDs and vice-versa.
2. Modify collector behavior to accommodate changes in datacollection groups by
recreating the RRD (if necessary) to include newly configured mibobjs in the group.