Add requirements-reference.txt for reproducible Python deps

[PranjalManhgaye]

Description

Problem
For each distribution release we need reproducible Python dependency versions, but pinning in every tutorial requirements.txt would require a lot of merge back-and-forth. There is also no explicit list of versions known to work for a given release.

Approach
Introduce a separate requirements-reference.txt (lockfile-style) with pinned versions, without changing the existing tutorial requirements.txt or run scripts. The reference file is only a manifest of versions that are known to work for a distribution.

What changed

• Add tools/tests/requirements-reference.txt with pinned versions (pyprecice, numpy, matplotlib, nutils, setuptools)
• Add update_requirements_reference.py to regenerate it from reference_versions.yaml (pyprecice from PYTHON_BINDINGS_REF)
• Add validate_requirements_reference.py to ensure the pyprecice version matches reference_versions.yaml
• Add GitHub Action check-requirements-reference.yml to run this validation on PR/push
• Document the mechanism in tools/tests/README.md and add a reminder to the release PR template

How to update at release

• Preferred: pip freeze from the systemtest Docker image, then pip freeze | python3 update_requirements_reference.py --from-freeze
• Or: python3 update_requirements_reference.py (uses reference_versions.yaml for pyprecice and script defaults for other packages)

Fixes #610

---

Checklist:

• I added a summary of any user-facing changes (compared to the last release) in the changelog-entries/<PRnumber>.md.
• I will remember to squash-and-merge, providing a useful summary of the changes of this PR.

[MakisH]

Thank you for contributing. I think the idea of having a script to generate requirements-reference.txt is indeed one possible solution to #610. However, I don't see the connection to reference_versions.yaml, which is about other dependencies than the requirements.txt.

Note that #610 mainly concerns the various Python-based tutorials that have requirements.txt. The system tests scripts are not that interesting for reproducibility: we mainly need the latest version to work. But when someone needs a older version of a Python-based tutorial, it is important that they can reproduce the original dependencies.

I think we need such a script to report the installed requirements.txt at a given time, on demand, plus some more version restrictions in the each requirements.txt. These version restrictions should state the requirements, without enforcing specific versions, so that the latest supported can generally be installed.

See how many tutorials require numpy >1 <2, or precice~=3.0.

Note that we should be able to get the latest available packages on PyPI, not just the ones that PIP might have cached on the local system.

Would you still like to continue working on this PR, or would you prefer to start a new one?

[PranjalManhgaye]

@MakisH for now i am contnuing on this pr only and work with your given direction ,
i am reworking it to focus on tutorial requirements.txt (looser constraints) plus an on-demand tools/report_tutorial_requirements.py that resolves versions from PyPI into root-level requirements-reference.txt , without tying it to reference_versions.yaml ,,

[PranjalManhgaye]

Closing in favor of a new PR with the reworked implementation (rebased onto latest develop, tutorial-centric approach per review) ,,