Add high-accuracy test runner script for generating and comparing CAMB references

[cmbant]

Motivation

• Provide a standalone test driver to generate high-accuracy reference ini files and outputs and to compare default runs against those references.
• Enable adjustable accuracy boosts and stability tuning for systematic high-precision validation of Fortran CAMB outputs.
• Allow both module-based and command-line execution and support for overriding ini parameters across generated tests.

Description

• Add fortran/tests/run_high_accuracy_tests.py, a new script that implements argument parsing and two main modes: generate_reference to create high-accuracy INI files and produce reference outputs, and compare to run standard tests and diff them against references using CAMB_test_files integration.
• Implement high-accuracy parameter adjustments via flags such as --int_tol_boost, --l_sample_boost, --l_accuracy_boost, --lens_potential_accuracy, --lens_margin, --stability_factor, and --do_late_rad_truncation, applied in _apply_high_accuracy_settings before writing reference INIs.
• Support overrides (--override KEY=VALUE), cleaning of target dirs, runner selection (--runner), and output customization; create reference INIs with make_reference_inis and execute them with run_reference_inis.
• Add numeric comparison and summary logic in summarize_differences which reads numeric tables, computes per-file and overall RMS(relative) and worst errors, and prints top offenders.

Testing

• No automated tests were added or executed as part of this change.

---

Codex Task

[cmbant]

Thanks, will try to have a look before too long.

Did self.assertAlmostEqual(pars.YHe, 0.246218, 5) give you a problem? Seems fine on the github runners.

The vscode config change also looks odd.

Does this method of configuring a BaseModel+ work with Cobaya?

@Auggie review