test: Update test tolerance for SciPy v1.16.x series

[kratsg]

Description

Resolves #2593.

• Update central value for test_asymptotic_calculator_has_fitted_par test to meet test tolerances, given numerical differences in SciPy v1.16.x from a rewrite of optimization routines from Fortran to C.
  • c.f. SciPy v1.16.0 gives different numerical results and Singular matrix E in LSQ subproblem #2593 (comment)
• Update doctests to use np.isclose to avoid having to revise exact numerical values.

Checklist Before Requesting Reviewer

• Tests are passing
• "WIP" removed from the title of the pull request
• Selected an Assignee for the PR to be responsible for the log summary

Before Merging

For the PR Assignees:

• Summarize commit messages into a comprehensive review of the PR

* Update central value for test_asymptotic_calculator_has_fitted_par test to
  meet test tolerances, given numerical differences in SciPy v1.16.x from a
  rewrite of optimization routines from Fortran to C.
   - c.f. https://github.com/scikit-hep/pyhf/issues/2593#issuecomment-3410710074
* Update doctests to use `np.isclose` to avoid having to revise exact numerical values.

[matthewfeickert]

unsure why things have changed in nightly

What's the environment diff?

[matthewfeickert]

unsure why things have changed in nightly

What's the environment diff?

This is because of the v1.16.3 patch release of SciPy that we requested (#2593 (comment)) to fix #2593. However, it was noted that there were numerical differences between Fortran and C given different compiler optimizations (#2593 (comment)).

It was also suggested that we should try to tune the pytest tolerances (#2593 (comment)) instead of changing the numerical values.

[kratsg]

This is because of the v1.16.3 patch release of SciPy that we requested (#2593 (comment)) to fix #2593. However, it was noted that there were numerical differences between Fortran and C given different compiler optimizations (#2593 (comment)).

It was also suggested that we should try to tune the pytest tolerances (#2593 (comment)) instead of changing the numerical values.

But it's already at rel=1e-3 and the different is larger than that. We can tune the central value to match both...

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 98.23%. Comparing base (2af4dea) to head (b4516fd).
⚠️ Report is 2 commits behind head on main.

Additional details and impacted files

@@           Coverage Diff           @@
##             main    #2631   +/-   ##
=======================================
  Coverage   98.23%   98.23%           
=======================================
  Files          65       65           
  Lines        4198     4198           
  Branches      592      592           
=======================================
  Hits         4124     4124           
  Misses         45       45           
  Partials       29       29           

Flag	Coverage Δ
contrib	97.97% <ø> (ø)
doctest	98.09% <ø> (ø)
unittests-3.10	96.28% <ø> (ø)
unittests-3.11	96.28% <ø> (ø)
unittests-3.12	96.28% <ø> (ø)
unittests-3.13	96.28% <ø> (ø)
unittests-3.8	96.28% <ø> (ø)
unittests-3.9	96.33% <ø> (ø)

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

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[matthewfeickert]

But it's already at rel=1e-3 and the different is larger than that. We can tune the central value to match both...

Note in #2593 (comment) the absolute difference tolerance is a the level of 1e-7.