Stabilize Merwe sigma-point scaling for small alpha

[FlorianPfaff]

Summary

• compute the Merwe sigma-point scale directly as alpha**2 * (n + kappa)
• avoid reconstructing it through n + (alpha**2 * (n + kappa) - n)
• reject non-finite or unrepresentable scales and weights with clear ValueErrors
• add a regression for a valid small positive alpha

Bug

MerweScaledSigmaPoints first formed

lambda_ = alpha**2 * (n + kappa) - n
scale = n + lambda_

For sufficiently small but representable positive scales, subtracting and re-adding n causes catastrophic cancellation. With n=2, kappa=0, and alpha=1e-9, the mathematically correct scale is 2e-18, but the current implementation rounds lambda_ to -2.0 and reconstructs scale as exactly 0.0. Construction then fails with division by zero, and sigma-point generation would collapse the spread.

Fix

Compute the positive scale directly and reuse that value for both weights and Cholesky scaling. This preserves the intended formula without the cancellation-prone round trip through lambda. The implementation also validates that the resulting scale and weights remain finite.

Validation

• reproduced the pre-fix values: legacy scale 0.0, direct scale 2e-18
• verified the direct scale yields finite weights and sigma offsets with radius sqrt(2e-18)
• added backend-aware regression coverage for finite weights and nonzero sigma-point spread
• python -m py_compile passes for the modified source and regression test
• branch is two commits ahead of main and zero behind

[github-actions]

✅MegaLinter analysis: Success

Descriptor	Linter	Files	Fixed	Errors	Warnings	Elapsed time
✅ COPYPASTE	jscpd	yes		no	no	21.57s
✅ JSON	prettier	7	0	0	0	0.95s
✅ JSON	v8r	7		0	0	4.46s
✅ MARKDOWN	markdownlint	68	0	0	0	1.8s
✅ MARKDOWN	markdown-table-formatter	68	0	0	0	0.81s
✅ PYTHON	black	1490	183	0	0	89.21s
✅ PYTHON	isort	1490	328	0	0	2.52s
✅ REPOSITORY	betterleaks	yes		no	no	1.99s
✅ REPOSITORY	checkov	yes		no	no	54.98s
✅ REPOSITORY	gitleaks	yes		no	no	14.48s
✅ REPOSITORY	git_diff	yes		no	no	0.22s
✅ REPOSITORY	secretlint	yes		no	no	66.38s
✅ REPOSITORY	syft	yes		no	no	5.15s
✅ REPOSITORY	trivy-sbom	yes		no	no	5.1s
✅ REPOSITORY	trufflehog	yes		no	no	31.13s
✅ YAML	prettier	11	0	0	0	0.67s
✅ YAML	v8r	11		0	0	11.05s
✅ YAML	yamllint	11		0	0	0.61s

Notices

📣 MegaLinter 9.5.0 is out! Discover the new features and security recommendations in the release announcement. (Skip this info by defining SECURITY_SUGGESTIONS: false)

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@9.6.0 --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_GITLEAKS,REPOSITORY_BETTERLEAKS,REPOSITORY_SECRETLINT,REPOSITORY_SYFT,REPOSITORY_TRIVY_SBOM,REPOSITORY_TRUFFLEHOG,YAML_PRETTIER,YAML_YAMLLINT,YAML_V8R

Show us your support by starring ⭐ the repository

[FlorianPfaff]

Closing as duplicate of #4289. The latter is the preferred Merwe sigma-point cancellation fix.