Fix PyTorch maximum/minimum device placement

[FlorianPfaff]

Summary

• Add a PyTorch backend-support hook for maximum and minimum that prefers an existing non-CPU tensor device before evaluating the elementwise operation.
• Wire the hook into the normal package import path so both raw pyrecest._backend.pytorch helpers and the active public pyrecest.backend facade are patched.
• Add subprocess regression coverage for raw and public PyTorch backend calls using CUDA when available and meta otherwise.

Bug fixed

pyrecest._backend.pytorch.maximum and minimum coerced their operands independently via array(...). Calls combining array-like or CPU operands with an existing CUDA/meta tensor could therefore end up invoking torch.maximum/torch.minimum on mixed-device tensors. The new hook moves both operands to the preferred existing tensor device and promotes to a common dtype before calling PyTorch.

Validation

• Inspected the branch diff against main (ahead_by=3, behind_by=0).
• Syntax-checked the new hook and regression test content locally with ast.parse.
• Smoke-checked PyTorch meta execution for torch.maximum/torch.minimum locally.
• Full repository test suite not run in this connector session; CI should run the focused regressions.

[github-actions]

✅MegaLinter analysis: Success

Descriptor	Linter	Files	Fixed	Errors	Warnings	Elapsed time
✅ COPYPASTE	jscpd	yes		no	no	74.47s
✅ JSON	prettier	7	0	0	0	0.92s
✅ JSON	v8r	7		0	0	5.24s
✅ MARKDOWN	markdownlint	68	0	0	0	1.94s
✅ MARKDOWN	markdown-table-formatter	68	0	0	0	0.88s
✅ PYTHON	black	1186	30	0	0	67.86s
✅ PYTHON	isort	1186	54	0	0	6.08s
✅ REPOSITORY	checkov	yes		no	no	48.86s
✅ REPOSITORY	gitleaks	yes		no	no	11.78s
✅ REPOSITORY	git_diff	yes		no	no	0.22s
✅ REPOSITORY	secretlint	yes		no	no	46.14s
✅ REPOSITORY	syft	yes		no	no	4.31s
✅ REPOSITORY	trivy-sbom	yes		no	no	7.32s
✅ REPOSITORY	trufflehog	yes		no	no	19.33s
✅ YAML	prettier	11	0	0	0	0.83s
✅ YAML	v8r	11		0	0	10.52s
✅ YAML	yamllint	11		0	0	0.56s

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@9.5.0 --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_SECRETLINT,REPOSITORY_SYFT,REPOSITORY_TRIVY_SBOM,REPOSITORY_TRUFFLEHOG,YAML_PRETTIER,YAML_YAMLLINT,YAML_V8R

Show us your support by starring ⭐ the repository

[FlorianPfaff]

Closing this conflicted PR as superseded by the main-branch PyTorch device compatibility hook repair in d4f7ba2.