Fix hyperhemisphere component state assignment

[FlorianPfaff]

Bug

HyperhemisphereCartProdParticleFilter had two inconsistent paths for assigning a single hyperhemispherical component distribution to a Cartesian-product particle filter.

The filter_state property setter correctly sampled n_particles * n_hemispheres component vectors before reshaping them into the product state. However, set_state(...) bypassed that path and sampled only n_particles vectors, then tried to construct a product distribution from the single-component-width array. For filters with more than one hemisphere this raises ValueError: Dimension is not correct. instead of assigning the distribution across all components.

The specialized filter_state setter also replaced particle locations without resetting particle weights. That differs from the base AbstractParticleFilter.filter_state contract, which resets weights to uniform whenever a new continuous distribution is sampled into the particle state. A prior nonuniform posterior could therefore leak stale weights into a newly sampled state.

Fix

• route non-product set_state(...) assignments through the existing product-aware filter_state setter;
• reset particle weights to uniform after sampling a hyperhemispherical/hyperspherical distribution, matching the base particle-filter assignment semantics;
• use backend ones_like(...) for the reset so weight dtype/device are preserved.

Regression coverage

Add focused NumPy-backend tests that verify:

• set_state(...) expands a single HyperhemisphericalWatsonDistribution across both product components without a shape error;
• the resulting product state has shape (n_particles, n_hemispheres * (dim_hemisphere + 1));
• samples remain on the upper hemisphere;
• both set_state(...) and direct filter_state = distribution assignment reset stale nonuniform weights to uniform.

Validation

• branch is based directly on current main (f196ec41d1d8681e279936db9f1a0a3abfa276e7);
• branch is 3 commits ahead / 0 behind;
• diff is limited to the hyperhemisphere Cartesian-product particle filter and one focused regression module;
• focused Ruff validation passed on the first PR head; GitHub Actions is authoritative for the full backend, packaging, documentation, security, and integration matrix.

[github-actions]

✅MegaLinter analysis: Success

Descriptor	Linter	Files	Fixed	Errors	Warnings	Elapsed time
✅ COPYPASTE	jscpd	yes		no	no	24.71s
✅ JSON	prettier	7	0	0	0	1.34s
✅ JSON	v8r	7		0	0	4.96s
✅ MARKDOWN	markdownlint	68	0	0	0	2.0s
✅ MARKDOWN	markdown-table-formatter	68	0	0	0	0.74s
✅ PYTHON	black	1867	88	0	0	95.02s
✅ PYTHON	isort	1867	141	0	0	2.79s
✅ REPOSITORY	betterleaks	yes		no	no	2.77s
✅ REPOSITORY	checkov	yes		no	no	56.51s
✅ REPOSITORY	gitleaks	yes		no	no	17.63s
✅ REPOSITORY	git_diff	yes		no	no	0.44s
✅ REPOSITORY	secretlint	yes		no	no	87.14s
✅ REPOSITORY	syft	yes		no	no	5.55s
✅ REPOSITORY	trivy-sbom	yes		no	no	5.6s
✅ REPOSITORY	trufflehog	yes		no	no	32.45s
✅ YAML	prettier	11	0	0	0	0.78s
✅ YAML	v8r	11		0	0	11.97s
✅ YAML	yamllint	11		0	0	0.67s

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