Refactor pedigree simulations to use tables allocated nodes

[jeromekelleher]

#1183 made some changes to the pedigree simulations, and the code is in a semi-working state. The new approach is to fully allocate all pedigree individuals and their nodes before the pedigree simulation is run. The simulation then initialises its state from this starting point, and then fills in the edge table with the simulated genetic relationships.

The code isn't currently using the nodes that were input, however, but is creating new nodes as they are needed. It should be a straightforward change to msp_merge_ancestors to use a pre-existing node rather than allocating a new one. There's various notes like this in msprime.c which should help to orient things.

We should also remove the special cases in the inner loops of merge_ancestors for pedigrees. We should be able to do this using the standard populations.

[jeromekelleher]

@DomNelson I think this is the main set of changes that we need to make to the underlying C code. Can you have a look and see how much is involved, please?

[jeromekelleher]

Closed in #1846 and earlier updates.