feat: add optimistix optimizer

[pfackeldey]

Description

This PR adds optimistix as a new optimizer backend. Optimistix implements JAX-native solvers and supports differentiation through minimization via the implicit function theorem, moving pyhf a step closer to being fully differentiable out of the box.

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

---

* Add new `optimistix_optimizer` class wrapping `optx.minimise` 
  with DFP solver by default (maxiter=1000); fixed parameters are 
  represented as `None` pytree leaves so the minimizer skips
  gradient updates for them; returns a `scipy.optimize.OptimizeResult`
  including `optx_state`
* Add `optimistix` optional dependency in `pyproject.toml` constrained
  to `python_version >= '3.11'`; wire into `backends`, `all`, and `test`
  pixi environments
* Extend `tests/conftest.py` `backend` fixture to conditionally include
  a `jax_optimistix` parametrize entry when `optimistix` is importable, so 
  all existing parametrized tests exercise the new optimizer automatically
* Add targeted tests:`ValueError` on `solver=None`, `ValueError` when
  used with non-JAX backend, basic fit, fit with a fixed parameter, 
  and `return_result_obj` round-trip checking `optx_state` is present

Co-Authored-By: Giordon Stark <kratsg@gmail.com>

Summary by CodeRabbit

• New Features
  • Added an Optimistix-based optimizer backend (Python 3.11+), including fixed-parameter handling and richer result metadata.
  • Added empty and arange array-creation utilities to tensor backends.
• Bug Fixes
  • Improved optional backend import errors when Optimistix is unavailable.
• Documentation
  • Expanded Sphinx cross-references to include Optimistix docs.
• Tests
  • Added/updated coverage and a new test marker related to the JAX+Optimistix combination.
• Chores
  • Updated optional dependency groups and tool configuration; improved local artifact ignore rules.

[lukasheinrich]

this looks ok to me.. just had minor comments..

[pfackeldey]

Ok, now it seems like I have to skip the optimistix tests for py<3.11 test suites. Do you have a suggestion where to apply such skips? I also see there's a constraint.txt in tests/ but I'm not sure if this is somehow related?

[codecov]

Codecov Report

❌ Patch coverage is 85.18519% with 12 lines in your changes missing coverage. Please review.
✅ Project coverage is 98.03%. Comparing base (826a041) to head (5f7ae6f).
⚠️ Report is 9 commits behind head on main.

Files with missing lines	Patch %	Lines
src/pyhf/tensor/numpy_backend.py	22.22%	7 Missing ⚠️
src/pyhf/tensor/jax_backend.py	66.66%	3 Missing ⚠️
src/pyhf/optimize/opt_optimistix.py	96.22%	1 Missing and 1 partial ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #2678      +/-   ##
==========================================
- Coverage   98.28%   98.03%   -0.25%     
==========================================
  Files          65       66       +1     
  Lines        4305     4386      +81     
  Branches      465      469       +4     
==========================================
+ Hits         4231     4300      +69     
- Misses         46       57      +11     
- Partials       28       29       +1     

Flag	Coverage Δ
contrib	97.92% <85.18%> (-0.24%)	⬇️
doctest	98.03% <85.18%> (-0.25%)	⬇️
unittests-3.10	94.91% <12.34%> (-1.56%)	⬇️
unittests-3.11	96.26% <85.18%> (-0.21%)	⬇️
unittests-3.12	96.26% <85.18%> (-0.21%)	⬇️
unittests-3.13	96.26% <85.18%> (-0.21%)	⬇️
unittests-3.9	94.98% <12.34%> (-1.56%)	⬇️

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

☔ View full report in Codecov by Harness.
📢 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.

[kratsg]

once tests are fixed and running, can go in.

[kratsg]

@pfackeldey it looks like we're seeing significantly longer run times for some of the tests with optimistix. If you look at the public_api tests, the hypotest we run with jax_optimistix takes significantly longer -- any ideas why?

[pfackeldey]

@kratsg thanks for pushing this PR forward!! I didn't have the capacity last week to do so.

I need to check, maybe a recompilation+retracing issue? I'll investigate...

[matthewfeickert]

The failures are due to the new matplotlib release (v3.11.0). We'll fix this and then rebase this.

[pfackeldey]

Once it's rebased the PR is ready from my side @matthewfeickert @kratsg :)