Make dependence on 0/1 mutations in ancestors ts explicit or remove

[jeromekelleher]

Currently we have to have 0/1 mutation in the ancestors ts, or assertions will fail. Either

1. Make this requirement explicit and test for it or
2. Update the input code to ignore the actual states of the values.

Probably (1) is less nasty, as we're unlikely to forget about this if we're forced to set the values to 0/1 all the time.

[jeromekelleher]

Also need to update make_ancestors_ts to allow us set these values to 0/1 appropriately.

[hyanwong]

I think #437 is probably relevant here: we should revisit this issue as part of allowing >2 alleles in matching.

[jeromekelleher]

I'm not sure this is true any more - we're matching with arbitrary alleles in the extend operation happily (if probably slightly wrongly #437)

[jeromekelleher]

I'm going to close - we can reopen if necessary.