feat(moore): link vLLM rms_norm

[voltjia]

Summary

• Add Moore vLLM rms_norm as linked implementation slot 16.
• Reuse TorchRmsNorm<Backend> for tensor conversion, contiguous staging, stream guards, and output copy-back.
• Cover CUDA and MUSA non-default streams through the existing C10<Device::Type::kMoore> bridge.

Motivation

This PR validates that the linked architecture works for a second operator on Moore. InfiniOps keeps rms_norm(input, weight, eps, out), while the provider ABI is:

voidrms_norm(at::Tensor& out, at::Tensor& input,
at::Tensor& weight, double epsilon);

The exact symbol was verified in the Moore vLLM 0.17.1 development image used by #874. Depends on #884. N/A - no linked issue.

Type of Change

• feat - new feature / new operator / new platform
• fix - bug fix
• perf - performance improvement (no behavioral change)
• refactor - code restructuring without behavior change
• test - adding or fixing tests only
• docs - documentation only
• build / ci - build system or CI configuration
• chore - tooling, formatting, or other non-code changes
• 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)
• PyTorch C++ bindings (WITH_TORCH)
• Build system / CMake / CI
• Python bindings / user-facing API

Smoke Test Result

Pending: operator-pruned Moore linked build, tests/test_rms_norm.py, and the existing smoke set.
The source upload to the Moore validation host requires explicit authorization.

Test Results on Supported Platforms

Platform	Affected	Build / Smoke Result	Full Result / Notes
NVIDIA	No	N/A - not affected	Host-independent clang-format/Ruff checks only
Iluvatar	No	N/A - not affected	N/A
MetaX	No	N/A - not affected	Separate draft PR
Cambricon	No	N/A - not affected	Separate draft PR
Moore	Yes	Pending remote linked build	Exact ABI verified in the #874 provider image
Ascend	No	N/A - not affected	N/A

Static validation

clang-format 21 --dry-run --Werror <changed C++ files>
exit 0
ruff 0.15.22 format --check tests/test_rms_norm.py
1 file already formatted
ruff 0.15.22 check tests/test_rms_norm.py
All checks passed!
git diff --cached --check
exit 0

Benchmark / Performance Impact

N/A - this PR validates linked integration correctness and makes no performance claim.

Notes for Reviewers

• Provider code contains no CUDA/MUSA preprocessor split; platform behavior remains in C10<Device::Type::kMoore>.
• TorchRmsNorm<Backend> preserves InfiniOps output-last ordering and stages non-contiguous tensors.
• This is the sixth PR in the linked-operator stack and depends on feat(metax): link vLLM rms_norm #884; its diff contains only the Moore provider and MUSA stream coverage.