Update KC metric to work on samples not leaves

[daniel-goldstein]

KC distance previously assumed that samples == leaves. However, the metric should measure the similarities of trees based on the relationships between the samples. This changes the metric to exclusively consider samples, not leaves. This means that internal samples no longer throw an exception and subtrees without any samples in them do not contribute to the KC distance.

It also makes more sense to focus on solely samples because the set of leaves is not consistent across trees in a tree sequence. Internal nodes that are roots in some trees and not in others affect how many leaves there are per tree. The sample set, however, is consistent across the sequence.

Resolves #575 #598

[codecov]

Codecov Report

Merging #610 into master will decrease coverage by 0.01%.
The diff coverage is 96.55%.

@@            Coverage Diff             @@
##           master     #610      +/-   ##
==========================================
- Coverage   87.31%   87.30%   -0.02%     
==========================================
  Files          22       22              
  Lines       16743    16733      -10     
  Branches     3274     3274              
==========================================
- Hits        14619    14608      -11     
  Misses       1040     1040              
- Partials     1084     1085       +1     

Flag	Coverage Δ
#c_tests	88.51% <96.55%> (-0.02%)	⬇️
#python_c_tests	90.46% <ø> (ø)
#python_tests	99.02% <ø> (ø)

Impacted Files	Coverage Δ
c/tskit/core.c	90.60% <ø> (-0.07%)	⬇️
python/tskit/trees.py	98.18% <ø> (ø)
c/tskit/trees.c	90.63% <96.55%> (-0.07%)	⬇️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 68400d8...e8e3bc5. Read the comment docs.

[jeromekelleher]

LGTM, thanks @daniel-goldstein! Minor comments above.

[jeromekelleher]

Looks great, thanks @daniel-goldstein!