Type changes at non-singleton nodes produce surprising output

[tgvaughan]

For instance, the following pair of reactions:

     C:1 -> C:1 + R:1
     C:1 + R:1 -> C:1

will produce a tree on which the R type never appears, despite the actual topology being correct.

The reason behind this is as follows. In output files, node types are used to refer to type types along the edges immediately above those edges. This is convention replicated internally, where each node in an inheritance graph is associated with exactly one population. Clearly, however, this creates problems when a node has an in-degree greater than 1, as there is no way of keeping track of those edge types.

You might imagine that this problem is insurmountable as we are using Newick as our output type, but in fact the extended Newick representation provides a single unique node for every edge in the network.

The fix is therefore to allow inheritance graph nodes to keep track of the types of all incoming edges rather than just the type at the node.

[tgvaughan]

This is actually quite difficult to solve, as edges aren't even uniquely identified by pairs of nodes - there may be more than one edge between a given node pair.