add soma function to return cell body positions

[jefferis]

[jefferis]

need to think about this a bit more. Does it just return xyz position of root node of neuron? Guess it could be overloaded e.g. to cope with catmaidneurons. Should it have a strict mode that requires unambiguous evidence that a node was tagged as the soma?

[jefferis]

• might need a special one for neuron lists that inspects attached data frame as well
• Some of this logic probably already exists in plot3d
• This could actually help with #304.
• We might want an assignment method as well – this could help with #206

[jefferis]

Further to this, there is at least one reason with catmaid neurons to want to separate the root node from the soma. The root node tells us where tracing started. This could also be specified with a tag of course, but this will not usually be the case.