allow new options to k-way stats

[petrelharp]

Over in #1623 we've added a new option, centre to the C k-way stats method because we wanted to add it to genetic_relatedness, and I couldn't see a way to add it to only genetic_relatedness. The option does nothing for other methods. This seems unfortunate, and led to having to do some hackey stuff in Fst, which calls both one- and k-way methods. We should figure out how to do this better, at least before we want to add any more arguments that affect only one method.

[jeromekelleher]

Would a solution be to use flags directly in the low-level Python C API? I.e., we export the flags values from _tskit, and pass an options argument through the low-level calls to enable different options. This will reduce boilerplate too, hopefully since there's less argument parsing in Python-C?

[petrelharp]

Oh, yes, that's the right way to do it. =) (I think - we'll have to try it to see!)

[jeromekelleher]

OK, let's do it as a follow up to #1623 - no point in complicating that with a bunch of infrastructure changes.

[jeromekelleher]

LMK if you want me to have a go at it.

[jeromekelleher]

OK if we push this one out to the 0.4.1 release @petrelharp? We're nearly there for 0.4.0 and it would be good to get it shipped.

[petrelharp]

Sure - I've got no bandwidth.