Poisson recombination in RecombCollector

[petrelharp]

Currently, RecombCollector

• requires simuPOP to have lociPos include 0.0 and the end of the chromosome
• when there's a recombination in an interval, just picks a point uniformly to cut instead of doing it Poisson

These are both maybe OK; especially the second thing would be annoying to do, because we'd need to consider inserting even numbers recombination points in intervals that simuPOP did not flag as having recombinations - if there are many loci this could be time consuming.