fix(mars): enable torch bindings and CONTRIBUTING smoke

[Ceng23333]

Summary

• Compile Mars torch unity sources with system g++ + -DUSE_HPCC=1 / HPCC includes (same MetaX/Moore pattern for vendor-forked torch headers).
• Add TorchDeviceName<Device::Type::kMars> → "cuda" in src/torch/device_.h and src/torch/mars/c10.h stream helpers.
• Extend scripts/generate_torch_ops.py (kMars in _DEVICE_TYPES / _C10_DEVICE_TYPES) and hand-written torch instantiations under src/torch/ops/*/.
• Ignore htcc sibling symlinks (*.cc.cu) in .gitignore.

Motivation

CONTRIBUTING smoke (pip install .[dev] + pytest -m smoke) needs Python bindings and Mars torch device mapping. Keep that plumbing out of the kernel PR (#942).

Depends on #942 (kernels) and #941 (platform).

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

Note: Mars / HPCC (WITH_MARS) — torch bindings / CONTRIBUTING smoke enablement.

Smoke Test Result

Host: MetaX X203 / HPCC, container infinilm-dev-refactor-dev (/opt/conda torch 2.8.0+mars, HPCC).
Branch feat/mars-torch-smoke (stack tip). INFINI_RT_ROOT = InfiniLM Mars integration prefix.

python -m pip install .[dev] --no-build-isolation --no-deps \
  --config-settings=cmake.define.INFINI_RT_ROOT=/workspace/InfiniLM/build/integration/mars/prefix \
  --config-settings=cmake.define.WITH_MARS=ON \
  --config-settings=cmake.define.INFINI_OPS_SMOKE_BUILD=ON
# Successfully built InfiniOps (cp310-linux_aarch64); install exit 0

python -m pytest tests -m smoke -q --devices cuda
# 92 passed, 35 skipped, 6235 deselected in 7.59s

Test Results on Supported Platforms

Platform	Affected	Build / Smoke Result	Full Result / Notes
NVIDIA	no	N/A - not affected
Iluvatar	no	N/A - not affected
MetaX	no	N/A - not affected
Cambricon	no	N/A - not affected
Moore	no	N/A - not affected
Ascend	no	N/A - not affected
Mars/HPCC (WITH_MARS)	yes	pip smoke + pytest -m smoke passed (92 passed, 35 skipped)	--devices cuda (Mars torch). Skips are N/A/env/other-backend or missing Mars ports (cast/cat/linear/matmul), not regressions.

Full `pytest` output (optional)

s......ssssssssss....ssssssss........................................... [ 62%]
......................ssss.......s..........                             [100%]
92 passed, 35 skipped, 6235 deselected in 7.59s

Benchmark / Performance Impact

N/A

Notes for Reviewers

• Depends on feat(mars): add native Mars operator specializations #942 and feat(mars): add WITH_MARS platform and htcc wrapper #941. Stack tip after kernels.
• Torch-only diff vs kernels: Ceng23333/InfiniOps@feat/mars-native-kernels...feat/mars-torch-smoke
• After feat(mars): add native Mars operator specializations #942 merges, retarget this PR base to master.
• Supersedes monolithic feat(mars): add native Mars kernels and htcc wrapper #939 smoke claim with the stacked tip evidence above.