Stabilize ellipsoidal positive-definiteness validation

[FlorianPfaff]

Summary

Fix a reproducible numerical validation bug in AbstractEllipsoidalBallDistribution for positive-definite shape matrices with extreme but representable axis scales.

Bug

The constructor checked positive definiteness by applying eigvalsh directly to the raw shape matrix. For a valid diagonal matrix whose variances balance between about 4.49e307 and 2.23e-308, NumPy's symmetric eigensolver rounds the tiny positive eigenvalues to zero. The constructor therefore rejects a strictly positive-definite matrix before the existing log-Cholesky volume computation can handle it.

This is reproduced by the existing regression:

• tests.distributions.test_ellipsoidal_ball_volume_stability::TestEllipsoidalBallVolumeStability::test_balanced_extreme_axes_avoid_intermediate_product_overflow

The NumPy 3.13 CI artifact from PR #5364 fails there with ValidationError: shape_matrix must be positive definite.

Fix

Validate positive definiteness after diagonal congruence scaling:

1. require strictly positive diagonal entries;
2. scale A as D^{-1/2} A D^{-1/2} using successive divisions;
3. run eigvalsh on the balanced, unit-diagonal matrix.

For positive diagonal D, this congruence preserves inertia, so the mathematical positive-definiteness criterion is unchanged. It avoids the avoidable dynamic-range loss in the eigensolver. Singular and indefinite matrices remain rejected.

Scope

• one production file changed
• 12 additions / 1 deletion
• branch is based directly on current main (baac3b1736bb8e81eafc1e3f69e10104636103b9)
• no overlap with open PR Fix circular integration, scalar coercion, and extreme aggregation regressions #5364

The existing stability and invalid-shape regression tests cover the changed behavior; no test-only weakening is included.

[github-actions]

✅MegaLinter analysis: Success

Descriptor	Linter	Files	Fixed	Errors	Max errors	Warnings	Elapsed time
✅ COPYPASTE	jscpd	yes		no		no	7.26s
✅ JSON	prettier	7	0	0		0	0.6s
✅ JSON	v8r	7		0		0	4.59s
✅ MARKDOWN	markdownlint	68	0	0		0	2.08s
✅ MARKDOWN	markdown-table-formatter	68	0	0		0	0.41s
✅ PYTHON	black	1988	15	0		0	67.47s
✅ PYTHON	isort	1988	27	0		0	2.96s
✅ REPOSITORY	betterleaks	yes		no		no	1.35s
✅ REPOSITORY	checkov	yes		no		no	37.74s
✅ REPOSITORY	git_diff	yes		no		no	0.13s
✅ REPOSITORY	secretlint	yes		no		no	109.71s
✅ REPOSITORY	syft	yes		no		no	3.14s
✅ REPOSITORY	trivy-sbom	yes		no		no	2.93s
✅ YAML	prettier	11	0	0		0	0.78s
✅ YAML	v8r	11		0		0	15.99s
✅ YAML	yamllint	11		0		0	1.02s

Notices

⚠️ Your configuration references items that have been removed from MegaLinter and are ignored: MAKEFILE_CHECKMAKE. See Removed linters to find their replacements.

See detailed reports in MegaLinter artifacts

Your project could benefit from a custom flavor, which would allow you to run only the linters you need, and thus improve runtime performances. (Skip this info by defining FLAVOR_SUGGESTIONS: false)

• Documentation: Custom Flavors
• Command: npx mega-linter-runner@10.0.0 --custom-flavor-setup --custom-flavor-linters PYTHON_BLACK,PYTHON_ISORT,COPYPASTE_JSCPD,JSON_V8R,JSON_PRETTIER,MARKDOWN_MARKDOWNLINT,MARKDOWN_MARKDOWN_TABLE_FORMATTER,REPOSITORY_CHECKOV,REPOSITORY_GIT_DIFF,REPOSITORY_BETTERLEAKS,REPOSITORY_SECRETLINT,REPOSITORY_SYFT,REPOSITORY_TRIVY_SBOM,YAML_PRETTIER,YAML_YAMLLINT,YAML_V8R

Show us your support by starring ⭐ the repository