fix: remove unused device placeholders

[voltjia]

Summary

• Remove unsupported kKunlun and kQy device placeholders from src/device.h, Python device parsing, Torch device naming, and Torch explicit instantiations.
• Remove Kunlun / Qy entries from wrapper and Torch codegen device lists.
• Update the ATen operator backend example so it only lists currently wired device types.

Motivation

InfiniRT is removing the unused kKunlun and kQy placeholders in InfiniTensor/InfiniRT#24. InfiniOps aliases infini::rt::Device, so keeping these enum references would break builds once InfiniOps is built against that InfiniRT branch.

Closes N/A

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

# Static checks on Windows checkout
rg -n "kKunlun|kQy|kunlun|qy|INFINI_DEVICE_KUNLUN|INFINI_DEVICE_QY" .
# no matches

git diff --check
# passed

python -m py_compile scripts/generate_wrappers.py scripts/generate_torch_ops.py
# passed

# Combined branch verification on remote nvidia host, using:
# - InfiniRT branch: fix/remove-unused-device-types (InfiniTensor/InfiniRT#24)
# - InfiniOps branch: fix/remove-unused-device-placeholders
# - container: accelerator-dev/nvidia:latest

cmake -S InfiniRT -B InfiniRT/build \
  -DCMAKE_INSTALL_PREFIX=/workspace/infinirt-prefix \
  -DWITH_CPU=ON \
  -DWITH_NVIDIA=ON \
  -DINFINI_RT_BUILD_TESTING=ON
cmake --build InfiniRT/build -j"$(nproc)"
ctest --test-dir InfiniRT/build --output-on-failure
cmake --install InfiniRT/build

python3 -m pip install . --no-build-isolation --no-deps \
  --config-settings=cmake.define.INFINI_RT_ROOT=/workspace/infinirt-prefix \
  --config-settings=cmake.define.WITH_CPU=ON \
  --config-settings=cmake.define.WITH_NVIDIA=ON \
  --config-settings=cmake.define.INFINI_OPS_SMOKE_BUILD=ON
python3 -m pytest tests -m smoke -q --devices cpu nvidia

InfiniRT CTest: 100% tests passed, 0 tests failed out of 8
InfiniOps smoke: 71 passed, 14 skipped, 12211 deselected in 17.50s

Test Results on Supported Platforms

Platform	Affected	Build / Smoke Result	Full Result / Notes
NVIDIA	Yes	smoke passed with InfiniRT PR #24 branch	Full not run
Iluvatar	No	N/A - not affected	Not run
MetaX	No	N/A - not affected	Not run
Cambricon	No	N/A - not affected	Not run
Moore	No	N/A - not affected	Not run
Ascend	No	N/A - not affected	Not run

Full `pytest` output (optional)

Running 85 items in this shard
..............ss..ss....ss...................................ssssss..ss. [ 84%]
.............                                                            [100%]
71 passed, 14 skipped, 12211 deselected in 17.50s

Benchmark / Performance Impact

N/A - this removes unused device placeholders and does not change kernels or algorithms.

Notes for Reviewers

This PR is intended to be merged with or after InfiniTensor/InfiniRT#24. The validation above builds InfiniOps against that InfiniRT branch to cover the cross-repo compatibility boundary.