feat(cambricon): add Add operator

[baominghelly]

Summary

• Add a Cambricon backend for Add in src/native/cambricon/ops/add/, including alpha, broadcasting, non-contiguous layouts, and contiguous fast-path execution.
• Support float16, bfloat16, float32, int32, and int64, and skip dtype combinations that torch.mlu cannot clone before backend execution.

Motivation

Extend canonical Add coverage to Cambricon while matching the current input + alpha * other interface and preserving the default alpha=1 overload. The implementation handles both contiguous tensors and broadcast/strided layouts used by the existing cross-platform Add test matrix.

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 (requires a ! in the Conventional Commits prefix or a BREAKING CHANGE: footer)

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

Focused Cambricon Add build and full operator test in infinicore-cambricon-dev-libaoming on cambricon218_01:

cmake --build /workspace/build/infiniops-refresh-add --target ops -j8
[7/7] Linking CXX shared module src/ops.cpython-310-x86_64-linux-gnu.so

/torch/venv3/pytorch/bin/python3.10 -m pytest -q tests/test_add.py --devices cambricon
84 passed, 48 skipped in 0.80s

The repository-wide default Cambricon smoke set was not run; validation was scoped to the affected operator.

Test Results on Supported Platforms

Platform	Affected	Build / Smoke Result	Full Result / Notes
NVIDIA	No	N/A - not affected	N/A
Iluvatar	No	N/A - not affected	N/A
MetaX	No	N/A - not affected	N/A
Cambricon	Yes	Focused Add-only build passed	tests/test_add.py: 84 passed, 48 skipped
Moore	No	N/A - not affected	N/A
Ascend	No	N/A - not affected	N/A

Full `pytest` output

...s..sss...s..sss...s..sss...s..sss...s..sss...s..sss...s..sss...s..sss [ 54%]
...s..sss...s..sss...s..sss...s..sss........................             [100%]
84 passed, 48 skipped in 0.80s

Benchmark / Performance Impact

N/A. This PR adds backend coverage and makes no performance claim.

Notes for Reviewers

• The fast path handles contiguous tensors with identical shapes; the general path handles broadcasting and arbitrary input/output strides.
• Device-aware DispatchFunc maps the runtime dtype to Cambricon native types.
• Floating-point alpha is converted to float for device execution; integral tensors use an integral alpha, as validated by the base operator.
• The added MLU skips cover torch.mlu cloning limitations for int16 and unsigned integer dtypes before the InfiniOps kernel is invoked.
• clang-format 21.1.2 was installed in the Cambricon container; clang-format --dry-run --Werror passed for the two added native files.