Collapsing collections of nodes in the network

[willGraham01]

One of the features that the API will need is the ability to "collapse" collections of nodes into a single node, and apply some sensible rule for handling the resulting connections of the "super-"node.

In practice, this will be necessary when users want to ask questions like "what is the distance from node A to region B", where region B is comprised of a number of individual, node-level areas of the brain. But the API should implement it in a general setting, because there is ultimately no reason not to.

rustworkx has an appropriate function for this, so we just need to design a sensible API and decide what we're going to do with the resulting graph - since the aforementioned function is an in-place change.

class Connections:

    def collapse_nodes(self, nodes: Container[int], weight_collapse_fn)

To decide:

• How should edges be collapsed by default? Sum of weights? Maximum weight?
• Is this done in-place (the .network attribute is directly modified)?
  • If so, do we want a .restore method or something to re-build the network from the original, lowest-level regions?
• What do we do about any metadata?
  • Gut instinct is to simply carry over any metadata.
  • When we have hierarchical data, we'd need to make note of the association between the assigned index of the new node and the corresponding region representing the collapse.