Fix for #90: SWC parser creates multiple soma sections.

[hernando]

This happened when neurites were connected to arbitrary soma samples.

[juanchopanza]

It looks OK to me, but I would prefer it if somebody else can also have a quick look.

[dnachbaur]

Looks reasonable to me w/o knowing too much about SWC

[hernando]

Just a minor comment. Notice that the current parser does two different things when the soma is given as a contour and as a point. When the soma is a point, the first order sections are not connected to that point (that would be wrong for several use cases: distance to soma computation, simple sphere + tubes rendering, ...). On the other hand, when the soma is described as a contour (or a tree), then the neurites are connected to the soma samples. This may not be always correct, but when that's the case the swc file should be fixed because that's how I interpret that case.

[juanchopanza]

@hernando I assume you're talking about the internal representation of morphologies. But how does this translate to SWC and H5 files written from a single point soma?

[hernando]

The SWC ans H5 are already not fully equivalent and it's not possible to to SWC->H5->SWC and get the original result (that's not a big deal since there's no SWC writer anyway).
With this commit, the behaviour for single point somas has not changed compared to before. In that case the first point of a first order section is the point which has the soma sample as parent, i.e. the section is not extended adding the soma point as the first one.