distinguish between subvolumes and subvolume parts

As of now, each Subvolume has exactly one geometry and is therefore assumed to be of a primitive geometry type such as a simplex or hypercube (which is true when refining hypercubes, simplices, prisms). However, this is not true for pyramids and one can think of other applications where the volume attached to a specific vertex is more complex. I suggest to define subvolumes on a more abstract level. Similar to the udg-interface, each abstract subvolume could hold multiple SubvolumeParts, which are again of a primitive type. Functions such as volume() could still be called on the Subvolume-object with internal rerouting to the parts.