feat(cambricon): add copy provider

[baominghelly]

Summary

• Add a CNNL-based Cambricon provider for Copy using cnnlCopy_v2.
• Extend the shared CNNL tensor descriptor helper for scalar tensors and add scalar/strided KV-cache Copy coverage.

Motivation

Cambricon currently has no native provider for the non-deprecated Copy operator. This prevents layout-preserving, broadcast, and strided-output copy paths from running through InfiniOps on MLU devices.

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

A Copy-only Cambricon build passed:

cmake -S /workspace/src/InfiniOps-pr-copy \
  -B /workspace/build/infiniops-pr-copy \
  -DWITH_CAMBRICON=ON \
  -DGENERATE_PYTHON_BINDINGS=ON \
  -DINFINI_OPS_OPS=copy \
  -DCMAKE_BUILD_TYPE=Release
cmake --build /workspace/build/infiniops-pr-copy --parallel 32

[100%] Built target ops

Test Results on Supported Platforms

Platform	Affected	Build / Smoke Result	Full Result / Notes
NVIDIA	No	N/A - not affected	N/A - not affected
Iluvatar	No	N/A - not affected	N/A - not affected
MetaX	No	N/A - not affected	N/A - not affected
Cambricon	Yes	Targeted build passed	42 targeted Copy tests passed
Moore	No	N/A - not affected	N/A - not affected
Ascend	No	N/A - not affected	N/A - not affected

Targeted pytest output

PYTHONPATH=/workspace/build/infiniops-pr-copy-site \
LD_LIBRARY_PATH=/workspace/build/infiniops-pr-copy-site/infini:/workspace/install/infini-rt/lib:/usr/local/neuware/lib64 \
/torch/venv3/pytorch/bin/python3.10 -m pytest -q tests/test_copy.py --devices cambricon

..........................................                               [100%]
42 passed in 0.25s

Benchmark / Performance Impact

N/A. This adds a previously missing Cambricon provider and does not replace an existing implementation.

Notes for Reviewers

• The provider supports scalar, broadcast, non-contiguous input, and strided output tensors, including KV-cache slice writes.
• CNNL execution uses the caller's queue and validates caller-provided workspace capacity.
• The shared descriptor update represents scalar tensors as a one-element CNNL array and uses the 64-bit descriptor interface.
• No deprecated RearrangeInfinilm provider is added.