Fix methanol GLOA benchmark summary generation

[bernalde]

Summary

• Package the benchmark summary generator as gdplib.benchmark_summary so the installed gdplib-benchmark entry point can generate summaries without importing a repo-root-only script.
• Keep generate_benchmark_summary_all.py as a compatibility wrapper around the packaged implementation.
• Add a solver-free regression test using a methanol gdpopt.gloa result JSON to cover summary generation from benchmark result folders.
• Considered Plan PyPI release workflow and Pixi platform support #104: this PR does not change release workflow, Pixi platform policy, or pixi.lock.

Tests run

• /home/bernalde/.pixi/bin/pixi run pytest tests/test_benchmark.py::test_generate_summary_uses_packaged_summary_module -v --tb=short - passed.
• /home/bernalde/.pixi/bin/pixi run pytest tests/test_methanol.py -v --tb=short - passed, 3 tests.
• /home/bernalde/.pixi/bin/pixi run pytest 'tests/test_module_imports.py::TestModelConstruction::test_model_construction[methanol]' -v --tb=short - passed.
• /home/bernalde/.pixi/bin/pixi run pytest tests/test_benchmark.py -v --tb=short - passed, 18 tests.
• /home/bernalde/.pixi/bin/pixi run test - passed, 282 passed and 1 skipped.
• /home/bernalde/.pixi/bin/pixi run lint - passed. The non-blocking flake8 --exit-zero step still reports existing repository warnings.
• /home/bernalde/.pixi/bin/pixi run gdplib-benchmark run --instances methanol --strategies gdpopt.gloa --timelimit 60 --solver-profile gams-baron --run-id issue70_methanol_gloa_final_verify - passed with 1 result row, 0 failures, summary generated. Solver result: optimal, objective -1743.4292381783366.

Notes about tests not run

• I did not run the full PR58 benchmark matrix; this PR was verified with the focused methanol GLOA benchmark path because the full matrix is solver-backed and much broader than issue Refactor methanol to fix benchmark issues #70.

Closes #70

[bernalde]

Follow-up on the methanol GLOA correctness question:

This PR fixes the benchmark summary import path so the installed gdplib-benchmark entry point can complete summary generation. It does not change the methanol model or GDPopt behavior.

I tested the focused methanol GLOA path against Pyomo/pyomo#3940, which fixes GLOA convex affine cut generation:

• Current GDPlib Pixi Pyomo 6.10.0, command /home/bernalde/.pixi/bin/pixi run gdplib-benchmark run --instances methanol --strategies gdpopt.gloa --timelimit 60 --solver-profile gams-baron --run-id issue70_methanol_gloa_final_verify
  • termination: optimal
  • iterations: 1
  • objective: -1743.4292381783366
• Fix GLOA convex affine cut slope Pyomo/pyomo#3940 loaded via PYTHONPATH=/tmp/pyomo-pr3940, command PYTHONPATH=/tmp/pyomo-pr3940 /home/bernalde/.pixi/bin/pixi run gdplib-benchmark run --instances methanol --strategies gdpopt.gloa --timelimit 60 --solver-profile gams-baron --run-id issue70_methanol_gloa_pyomo3940 --no-summary
  • termination: optimal
  • iterations: 15
  • objective: -1793.4292385308854

The patched Pyomo result matches the documented methanol best-known objective -1793.4292381783 to solver tolerance. That strongly suggests the remaining methanol GLOA global-solution mismatch in #70 is a Pyomo GLOA affine-cut bug rather than a GDPlib methanol formulation bug.

Implication for this PR: #128 is still useful because it makes the benchmark CLI complete successfully after a methanol GLOA run, but we should not claim the current GDPlib dependency stack fixes GLOA optimality until the Pyomo fix is merged and available in the environment used for verification.

[bernalde]

Objective-sense note (context: Pyomo/pyomo#3986, merged 2026-08-13, fixed the GDPopt LOA objective-sense bug tracked in #114):

No changes needed here — this PR's scope is GLOA benchmark summary generation, and the fix was LOA-only (GLOA passes the objective sense through correctly), so the verification evidence stands. The related follow-up — reverting methanol's minimize(-profit) workaround to native maximize(profit) and updating the README note — is tracked in #70 for after a Pyomo release containing the fix, and is intentionally out of scope for this PR.