fix: align device list with InfiniRT master

[voltjia]

Summary

• Remove hard-coded kKunlun and kQy device references from active device lists, PyTorch backend instantiations, wrapper codegen, and docs examples.
• Align generated wrapper device mapping with the device enum exported by current InfiniRT master.

Motivation

Current InfiniOps master fails to build against current InfiniRT master because InfiniOps still references Device::Type::kKunlun and Device::Type::kQy, which are no longer exported by InfiniRT master. This blocks NVIDIA smoke builds with WITH_TORCH=ON, including the reduced PyTorch backend operator set.

Closes #

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

# ssh nvidia, accelerator-dev/nvidia:latest
# InfiniRT master: 0f0d61f
# InfiniOps branch: fix/infinirt-master-build

python3 -m py_compile scripts/generate_torch_ops.py scripts/generate_wrappers.py
git diff --check

scripts/dev/build.sh nvidia --smoke --jobs 4 --torch-jobs 1 --binding-jobs 1
...
-- Wrapper op allowlist: add;mul;cast;cat;gemm;matmul;linear;rms_norm;swiglu;causal_softmax;abs;clamp;exp
-- Codegen schema: PyTorch v2.10.0a0
-- Codegen torch op allowlist: abs;clamp;exp
generated 4 overloads across 3 ops: ['abs', 'clamp', 'exp']
-- Generating torch op wrappers - done
-- Torch unity build: 5 sources batched into 1 translation units
-- Generating wrappers - done
[11/12] Linking CUDA shared module src/ops.cpython-312-x86_64-linux-gnu.so
[build] done

PYTHONPATH=/workspace/InfiniOps/build-nvidia/install python3 -m pytest tests -m smoke -q --devices nvidia --tb=short
Running 56 items in this shard
......ssss..ss.............................ssssss.......                 [100%]
44 passed, 12 skipped, 6859 deselected in 8.41s

Test Results on Supported Platforms

Platform	Affected	Build / Smoke Result	Full Result / Notes
NVIDIA	Yes	smoke passed	Full suite not run; smoke build and smoke tests passed in accelerator-dev/nvidia:latest.
Iluvatar	No	N/A - not affected	Device enum alignment only; not run locally.
MetaX	No	N/A - not affected	Device enum alignment only; not run locally.
Cambricon	No	N/A - not affected	Device enum alignment only; not run locally.
Moore	No	N/A - not affected	Device enum alignment only; not run locally.
Ascend	No	N/A - not affected	Device enum alignment only; not run locally.

Full `pytest` output (optional)

Running 56 items in this shard
......ssss..ss.............................ssssss.......                 [100%]
44 passed, 12 skipped, 6859 deselected in 8.41s

Benchmark / Performance Impact

N/A - build compatibility fix only.

Notes for Reviewers

This PR intentionally follows the device enum currently exported by InfiniRT master. It does not add replacement support for removed kunlun or qy device names.

[voltjia]

Closing as duplicate of #790. The patch content is identical, and #790 has already been merged into master.