Stabilize SO(3) chordal smoother weights across backends

[FlorianPfaff]

Bug

SO3ChordalMeanSmoother._normalize_weight_vector() scales weights as weights / max(weights) on the active backend. That is numerically unsafe on JAX/XLA for two already-established PyRecEst edge cases:

• division by a near-maximum finite float can be lowered through a reciprocal that underflows, collapsing valid weight ratios;
• positive or negative subnormal weights can be flushed during backend arithmetic/comparisons, causing valid ratios to be rejected or a tiny negative weight to evade validation.

This is the same numerical class previously hardened in Dirac/mixture/particle-weight paths, but the SO(3) chordal smoother still had its own vulnerable normalization.

Fix

• validate finiteness/sign from the stored host representation so subnormal signs are preserved;
• scale weights to O(1) on the host before moving the ratios back to the active backend;
• normalize only the already-scaled ratios, avoiding reciprocal underflow at extreme magnitudes;
• retain the existing public error contract for non-finite, negative, and zero-total weights.

Regression coverage

Adds cross-backend tests for:

• a 2:1 ratio at the active dtype's largest finite value through the public chordal-mean path;
• a 2:1 ratio made from positive subnormals;
• rejection of a negative subnormal weight.

Scope is limited to the SO(3) chordal mean smoother and its focused tests.

[github-actions]

✅MegaLinter analysis: Success

Descriptor	Linter	Files	Fixed	Errors	Max errors	Warnings	Elapsed time
✅ COPYPASTE	jscpd	yes		no		no	7.67s
✅ JSON	prettier	7	0	0		0	0.78s
✅ JSON	v8r	7		0		0	4.4s
✅ MARKDOWN	markdownlint	68	0	0		0	2.15s
✅ MARKDOWN	markdown-table-formatter	68	0	0		0	0.41s
✅ PYTHON	black	2021	4	0		0	68.79s
✅ PYTHON	isort	2021	5	0		0	3.08s
✅ REPOSITORY	betterleaks	yes		no		no	1.45s
✅ REPOSITORY	checkov	yes		no		no	38.1s
✅ REPOSITORY	git_diff	yes		no		no	0.14s
✅ REPOSITORY	secretlint	yes		no		no	112.83s
✅ REPOSITORY	syft	yes		no		no	3.78s
✅ REPOSITORY	trivy-sbom	yes		no		no	2.77s
✅ YAML	prettier	11	0	0		0	0.8s
✅ YAML	v8r	11		0		0	17.7s
✅ YAML	yamllint	11		0		0	0.72s

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