Fix PyTorch symmetric Leopardi collar dtype

[FlorianPfaff]

Summary

Fix the PyTorch regression in symmetric Leopardi small-grid partitioning.

The existing regression

tests/test_leopardi_symmetric_small_grid.py::test_symmetric_four_point_partition_uses_two_collars

fails in the current Python 3.13 PyTorch CI matrix: get_equal_area_caps(2, 4, symmetric=True) returns region counts [3, 1, 1, 1] instead of the intended [1, 1, 1, 1].

Root cause

PR #5202 fixed small symmetric partitions by making the half-collar count an explicitly integer backend array:

array((1, round(ratio_half)), dtype=int32)

The later compatibility wrapper in src/pyrecest/sampling/leopardi_sampler/__init__.py reimplemented that logic but omitted dtype=int32. On PyTorch the tuple mixes the Python integer lower bound with a floating backend result from round(ratio_half), so the construction is promoted away from the discrete collar-count contract.

Fix

Restore the explicit backend int32 dtype in the compatibility wrapper, matching the canonical implementation already present in leopardi_sampler.py and the fix originally merged in #5202.

No partition formulas or public API are changed.

Evidence

• Reproduced from the existing PyTorch 3.13 test-results artifact of workflow run 32406079884 on a current-main-based branch.
• The focused regression already exists in the repository, so no duplicate test is added.
• Branch is based directly on main at b7d3edc9ac05600de5345dc20fd6f3db33154d86.
• Final diff: one production file, 4 additions / 1 deletion.
• No open Leopardi PR existed at the time of this fix.

[github-actions]

✅MegaLinter analysis: Success

Descriptor	Linter	Files	Fixed	Errors	Max errors	Warnings	Elapsed time
✅ COPYPASTE	jscpd	yes		no		no	7.98s
✅ JSON	prettier	7	0	0		0	0.64s
✅ JSON	v8r	7		0		0	5.13s
✅ MARKDOWN	markdownlint	68	0	0		0	2.06s
✅ MARKDOWN	markdown-table-formatter	68	0	0		0	0.44s
✅ PYTHON	black	1989	15	0		0	65.87s
✅ PYTHON	isort	1989	28	0		0	2.95s
✅ REPOSITORY	betterleaks	yes		no		no	1.22s
✅ REPOSITORY	checkov	yes		no		no	35.59s
✅ REPOSITORY	git_diff	yes		no		no	0.14s
✅ REPOSITORY	secretlint	yes		no		no	105.67s
✅ REPOSITORY	syft	yes		no		no	2.75s
✅ REPOSITORY	trivy-sbom	yes		no		no	3.19s
✅ YAML	prettier	11	0	0		0	0.64s
✅ YAML	v8r	11		0		0	17.19s
✅ YAML	yamllint	11		0		0	1.1s

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