fix(tests): generate integer inputs portably

[voltjia]

Summary

• Make randint_strided fall back to torch.randint(...).copy_() when device-specific random_ is unavailable.
• Preserve the existing fast path for devices that support random_ directly.

Motivation

Some PyTorch accelerator forks do not implement random_ for every integer dtype. The test utility should still be able to generate integer inputs portably across supported platforms.

Closes # N/A

Type of Change

• N/A feat — new feature / new operator / new platform
• fix — bug fix
• N/A perf — performance improvement
• N/A refactor — code restructuring without behavior change
• N/A test — adding or fixing tests only
• N/A docs — documentation only
• N/A build / ci — build system or CI configuration
• N/A chore — tooling, formatting, or other non-code changes
• N/A Breaking change

Platforms Affected

• CPU (WITH_CPU)
• NVIDIA (WITH_NVIDIA)
• Iluvatar (WITH_ILUVATAR)
• MetaX (WITH_METAX)
• Cambricon (WITH_CAMBRICON)
• Moore (WITH_MOORE)
• Ascend (WITH_ASCEND)
• N/A PyTorch C++ bindings (WITH_TORCH)
• N/A Build system / CMake / CI
• N/A Python bindings / user-facing API

Test Results on Supported Platforms

Direct profile, PYTEST_WORKERS=1.

Platform	Built	pytest Result	Notes / Hardware
NVIDIA	Yes	4151 passed, 1375 skipped in 294.91s	Full pytest tests/. Matches #604/#605/#608 pass/skip counts.
Iluvatar	Yes	3651 passed, 375 skipped in 261.93s	Full pytest tests/. Matches #604/#605/#608 pass/skip counts.
MetaX	Yes	5795 passed, 1447 skipped in 361.78s	Full pytest tests/. Matches #604/#605/#608 pass/skip counts.
Cambricon	Yes	3073 passed, 3857 skipped in 892.41s	Full pytest tests/. Fixes the previous 12 tests/test_add.py integer-input failures seen in #604/#605/#608.
Moore	Yes	300 failed, 5459 passed, 1483 skipped in 541.04s	Full pytest tests/. Same known pre-existing tests/test_gemm.py MUSA failures as #604/#605/#608.
Ascend	Yes	3828 passed, 138 skipped in 455.21s; container exit code 137 after pytest	Full pytest tests/. No pytest failures in this run; same post-test exit-code behavior as #605/#608 and no regression versus #604.

Compared with the last merged baseline PRs #604, #608, and #605, this PR has no regression in build status or collected coverage. Cambricon improves by removing the 12 known integer-input failures from tests/test_add.py; the remaining known Moore failure class is unchanged.

Full validation summaries

ruff:
1 file already formatted
All checks passed!

nvidia build=0 test=0
4151 passed, 1375 skipped in 294.91s

iluvatar build=0 test=0
3651 passed, 375 skipped in 261.93s

metax build=0 test=0
5795 passed, 1447 skipped in 361.78s

cambricon build=0 test=0
3073 passed, 3857 skipped in 892.41s

moore build=0 test=1
300 failed, 5459 passed, 1483 skipped in 541.04s

ascend build=0 test=137
3828 passed, 138 skipped in 455.21s

Benchmark / Performance Impact

N/A. The fallback only runs when the direct device random_ path is unavailable.

Notes for Reviewers

The fallback copies values generated as int64 into the requested dtype/device, which keeps the helper usable on accelerator forks with incomplete integer RNG coverage.

---

Checklist

Title, Branch, and Commits

• PR title follows Conventional Commits.
• Branch name follows <type>/xxx-yyyy-zzzz.
• Each commit message follows Conventional Commits.
• Small PR is a single squashable commit.
• No stray merge commits from master.
• No fixup! / squash! / wip commits remain.

Scope and Design

• Changes are minimal.
• No dead code, debug prints, or unrelated TODOs were added.
• No unrelated formatting churn.
• N/A Public API changes.

General Code Hygiene

• Code is self-explanatory.
• Modified file ends with a newline.
• No trailing whitespace, tab/space mixing, or stray BOMs.
• N/A Comments/error messages with identifiers.
• N/A New comments.
• N/A New error messages.

C++ Specific

N/A. No C++ files changed.

Python Specific

• Code follows the surrounding Python style.
• ruff check tests/utils.py passed.
• ruff format --check tests/utils.py passed.
• N/A New comments.
• Framework-specific exception handling conventions are preserved.
• No blank line rule violations were introduced.
• Return spacing follows surrounding style.
• N/A Docstrings.
• N/A Type hints.

Testing

• Full platform testing was run across NVIDIA, Iluvatar, MetaX, Cambricon, Moore, and Ascend.
• N/A No new tests were added.
• N/A No test parameterization changes.
• N/A No new flaky tests.
• Regression coverage: Cambricon integer-input failures in tests/test_add.py are covered by full-platform validation and now pass.

Build, CI, and Tooling

• Fresh package build passed on all supported platforms.
• N/A compile_commands.json.
• N/A Backend auto-detection.
• N/A CUDA-like mutual exclusion.
• N/A CI workflows are expected to validate formatting.
• N/A Runtime dependencies.

Documentation

• N/A User-facing documentation changes.
• N/A New public utilities.
• N/A Breaking changes.

Security and Safety

• No secrets or personal data are committed.
• N/A Third-party code.
• N/A Unsafe pointer arithmetic.

[voltjia]

请 @bitzyz 初审，@Ziminli 终审。