Isolate JET and repair CI environments

[Krastanov]

Summary

• move JET into test/projects/jet and route nightly CI through the general/jet test_args matrix
• declare QuantumExpanders in the ordinary test workspace so package tests work on supported runners
• keep docs and test projects in the workspace while temporarily hiding the workspace from the downgrade resolver
• enforce the Julia 1.12-compatible direct floors: Graphs 1.12, PrecompileTools 1.3, Nemo 0.52.1, and Oscar 1.5
• remove the Windows test row because the required Polymake dependency explicitly does not support Windows; retain Ubuntu and Apple Silicon macOS
• skip documentation and changelog checks for Dependabot pull requests
• leave the benchmark reminder workflow intact

Validation

• the exact julia-downgrade-compat@v2 forcedeps resolution passes every declared direct floor on Julia 1.12
• Oscar 1.5 and its resolved Polymake/CxxWrap stack fully precompile and import on Julia 1.12
• strict Pkg.test(; test_args=["general"], allow_reresolve=false, force_latest_compatible_version=false) constructs the test sandbox while preserving the forced direct versions, admits IGraphs 1.0 with Graphs 1.12, and reaches the TestItemRunner workload
• one coverage run hit Aqua's ten-second persistent-task assertion, but focused Aqua.has_persistent_tasks probes at both tmax=10 and tmax=60 returned false (no persistent task); the timeout was not reproducible and the strict run continued into the known long Quantum Tanner workload
• the dedicated JET environment activates through Pkg.test(; test_args=["jet"])
• gh act -n --validate -W .github/workflows/ci.yml
• workflow YAML and TOML parse; git diff --check

The fresh PR checks are authoritative for full-suite completion.

[Krastanov]

Pardon the slop grenade https://noslopgrenade.com/ -- this is from an automated runner that I have tasked with fixing some menial CI issues. It did not do a great job here.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 95.52%. Comparing base (bef39e4) to head (beb61e4).

Additional details and impacted files

@@           Coverage Diff           @@
##             main      #87   +/-   ##
=======================================
  Coverage   95.52%   95.52%           
=======================================
  Files           8        8           
  Lines         760      760           
=======================================
  Hits          726      726           
  Misses         34       34           

☔ 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.

[Krastanov]

@Fe-r-oz , it seems most tests are failing, here and on master. I think the changes here are all good. Do you understand the source of the failures either on master or here? How do you safely develop this given that most tests seem to fail?

[Fe-r-oz]

Thank you for taking care of this, Stefan. I'll dig into the failures and get back to you. My first guess is the old random_quantum_tanner_code, which triggers a segfault when it runs in the test suite. I still need to track down why.

[Krastanov]

If it is unused and you are deeply certain it is not important for the correctness of your code, we can delete that part of the code. I will also ask one of my slop generators to look into this, maybe they will generate something useful.

[Krastanov]

It seems codex sol 5.6 figured out most of the fixes -- they will be added to a new commit that we can look at, to see whether we are happy with it.

[Krastanov]

@Fe-r-oz , this seems to fix the issues with the CI. Please check it out and merge if you agree