Use jukes cantor instead of msprime for failing test

[benjeffery]

Getting this error on Ci since msprime 1.0.0:

FAILED tests/test_tree_stats.py::TestSiteTajimasD::test_wright_fisher_simplified
E           AssertionError: 
E           Not equal to tolerance rtol=1e-07, atol=1e-06
E           
E           x and y nan location mismatch:
E            x: array([[nan, nan, nan, nan]])
E            y: array([[ nan,  nan,  nan, -inf]])

../../.local/lib/python3.7/site-packages/numpy/testing/_private/utils.py:746: AssertionError

Hopefully using the tskit jukes cantor fixes this.

[benjeffery]

I'm merging this now to get things moving, but perhaps @hyanwong or @petrelharp could sanity check my params.

[hyanwong]

Looks fine to me, especially as this is just for the test case. I was slightly thrown by the fact that you are using 5 "loci" in the WF model (i.e. 4 discrete recombination points) but 10 loci for the mutations, but I guess it doesn't really matter.

[codecov]

Codecov Report

Merging #1330 (0f18da6) into main (4ecb8cb) will not change coverage.
The diff coverage is n/a.

@@           Coverage Diff           @@
##             main    #1330   +/-   ##
=======================================
  Coverage   93.84%   93.84%           
=======================================
  Files          26       26           
  Lines       22273    22273           
  Branches     1029     1029           
=======================================
  Hits        20902    20902           
  Misses       1339     1339           
  Partials       32       32           

Flag	Coverage Δ
c-tests	92.44% <ø> (ø)
lwt-tests	92.97% <ø> (ø)
python-c-tests	95.18% <ø> (ø)
python-tests	98.89% <ø> (ø)

Flags with carried forward coverage won't be shown. Click here to find out more.

---

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 4ecb8cb...0f18da6. Read the comment docs.

[jeromekelleher]

The difference is technically a bug in the test suite - we can return -inf, nan or +inf from Tajima's D, depending. Our test version of the stat isn't catching this. I hit the same issue here: https://github.com/pystatgen/sgkit/pull/542