add export_recombination_node_breakpoints

[hyanwong]

Replaces #116, as discussed. Returns one row for each breakpoint. We can easily subset down to rows with only a single breakpoint (i.e. 2 parents) by

df.drop_duplicates('node', keep=False)

or average the metrics across all the breakpoints in a node by

df.groupby('node').mean()

So I think this is a nice and flexible output.

[hyanwong]

I did the reuse-the-tree thing, which is much faster, thanks for the hint @jeromekelleher - we should have a "tips" tutorial for things like this. But it turned out to be just as easy to make two independent tree objects at the start and seek each one to adjacent positions, rather than use one tree and keep all the parent node IDs for comparison in a list. I think this is fine.

[hyanwong]

Here's an example of the output