fix: real-valued high BC anisotropy on NumPy >= 2

[cbueth]

Description

Scheduled CI on main failed in the Partitioning and Metrics jobs: raised RepresenterError on values like (5.73+0j), and the anisotropy special-case tests got inf+nanj instead of inf.

There was a upstream change for NumPy >= 2, where np.linalg.eigvals returns complex128 eigenvalues even for real symmetric covariance matrices. This leaked complex scalars into the metrics and key figures.

This PR switches the calculation to np.linalg.eigvalsh (always real for symmetric input), clips numerical-noise negative eigenvalues, and returns a plain float. As a fallback, _make_yaml_compatible now collapses zero-imaginary complex scalars to float, so YAML dumps can't crash on a stray complex value again.

Type of Change

• 🐛 Bug fix (non-breaking change that fixes an issue)
• ✨ New feature (non-breaking change that adds functionality)
• 💥 Breaking change (fix or feature that would cause existing functionality to change)
• 📝 Documentation update
• 🧹 Code refactor (no functional changes)
• ✅ Test update

How Has This Been Tested?

• Unit tests
• Integration tests
• Manual testing

Test configuration:

• OS: MacOS, NumPy 2.5.2
• pytest "tests/metrics/test_metric.py::TestMetric::test_saving_and_loading" -k "ResidentialPart or BetweennessPart"
• pytest tests/partitioning/test_base.py -k "save_key_figures and Adliswil"

Checklist

• My code follows the project's style guidelines
• I have performed a self-review of my own code
• I have commented my code, particularly in hard-to-understand areas
• I have updated the documentation
• My changes generate no new warnings
• I have added tests that prove my fix/feature works
• New and existing tests pass locally

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!