null chromosomes should not be in the tree sequence

[bhaller]

This issue fills in for the part of #205 that is about not putting the null genomes into the tree sequence. The rest of #205 will be fixed now (making SLiM use null genomes instead of empty genomes when appropriate, such as in haploids), but this final aspect will be fixed later if at all. The reason is that the benefits of this last step are less clear, and the fix is also less clear because there are a number of places in SLiM where we assume two nodes per individual - search for * 2 + 1 to find at least some of them. @petrelharp

[bhaller]

This appears to have a real-world consequence within SLiM: checkCoalescence() does not function correctly for models containing null genomes, I think. See https://groups.google.com/g/slim-discuss/c/EhV-mjf_2Ls/m/-FoNMWVWAAAJ.