KC Distance seg faults on non-sample leaves

[daniel-goldstein]

Likely stems from the synonymous use of leaf and sample everywhere, which is not very consistent or correct. Related to #575

[jeromekelleher]

I think the right thing to do here is to change our definition of KC to be in terms of samples instead of leaves. Should be straightforward enough.

[daniel-goldstein]

Fixed by #610