Numerical stability of beta model

[JereKoskela]

Fixing a bias in Beta-coalescent TMRCAs due to numerical issues with acceptance probabilities.

Closes #2256

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Comparison is base (804e036) 98.70% compared to head (dcf123a) 91.52%.

❗ Current head dcf123a differs from pull request most recent head fdc169c. Consider uploading reports for the commit fdc169c to get more accurate results

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #2257      +/-   ##
==========================================
- Coverage   98.70%   91.52%   -7.19%     
==========================================
  Files          11       20       +9     
  Lines        4022    11337    +7315     
  Branches      907     2304    +1397     
==========================================
+ Hits         3970    10376    +6406     
- Misses         28      523     +495     
- Partials       24      438     +414     

Flag	Coverage Δ
C	91.52% <100.00%> (?)
python	98.70% <ø> (ø)

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

Files	Coverage Δ
lib/msprime.c	86.12% <100.00%> (ø)

... and 8 files with indirect coverage changes

---

Continue to review full report in Codecov by Sentry.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 804e036...fdc169c. Read the comment docs.

[jeromekelleher]

LGTM. Can you update the CHANGELOG please to note this bugfix?

Shall we push out a quick bugfix release?

[JereKoskela]

I'm guessing that the CHANGELOG should have an entry along these lines:

Bug fixes:

Change tolerance of polynomial approximation in Beta-coalescent acceptance probabilities ({issue}2256, {pr}2257, {user}JereKoskela)

But what should the header be? All existing entries are written under version numbers of the corresponding release, and this one doesn't have one (at least yet). I'm happy for you to enter this in too if that's quicker than explaining the process to me @jeromekelleher.

As for whether to release a bugfix, I have no idea what that entails so happy to leave the decision to you. I doubt this is having a big impact and waiting a little while is unlikely to cause havoc if it would be simpler to wait until a more substantial release is ready.

[GertjanBisschop]

It is as simple as providing a provisional header @JereKoskela.
[1.3.1] - 2024-XX-XX

[jeromekelleher]

Perfect, thanks @JereKoskela

[jeromekelleher]

OK, maybe just squash those commits now @JereKoskela and we'll merge.

I think we'll just pop out a quick release. This is the only change and it's easy to do. Are you happy to shepherd the release process @GertjanBisschop ?

[JereKoskela]

Squash done. Over to you @jeromekelleher and @GertjanBisschop