Fix bug in missing ancestral states

[hyanwong]

A silly typo ("ancestral_allele" -> "ancestral_state") revealed that the test suite wasn't comprehensive enough for missing ancestral states, and needs a bit of tweaking to be able to test this corner case.

[codecov]

Codecov Report

Merging #738 (e43d749) into main (de277cf) will not change coverage.
The diff coverage is n/a.

@@           Coverage Diff           @@
##             main     #738   +/-   ##
=======================================
  Coverage   93.36%   93.36%           
=======================================
  Files          17       17           
  Lines        5317     5317           
  Branches      977      977           
=======================================
  Hits         4964     4964           
  Misses        233      233           
  Partials      120      120           

Flag	Coverage Δ
C	93.36% <ø> (ø)
python	96.54% <ø> (ø)

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

Impacted Files	Coverage Δ
tsinfer/inference.py	98.75% <ø> (ø)

📣 We’re building smart automated test selection to slash your CI/CD build times. Learn more

[hyanwong]

It would be good to merge this bugfix: the codebase change is minimal, although the test suite change is more involved, sadly. I have simply called return from TestSparseAncestorsRoundTrip:: verify_data_round_trip for the newly introduced case where we have unknown ancestral states (and therefore can't generate sties without placing some on using parsimony), but perhaps I should call some form of skip instead?

[jeromekelleher]

The test suite changes take a bit of time to digest I'm afraid, looking at it now.

[hyanwong]

Yes, sorry about that. It would probably be easier if we had #739 , but I though it better to fix ASAP.

[jeromekelleher]

LGTM