Refactor: collapse ChipWorker init/set_device → init(device_id, bins)

[ChaoWao]

Summary

ChipWorker exposed init / set_device / reset_device / finalize as four public lifecycle methods even though set_device existed only to satisfy CANN's per-thread aclrtSetDevice attach requirement (#715), and reset_device had zero call sites. The split was already the source of one bug (#715) and forced every caller through a redundant ceremony.

This PR collapses the lifecycle to a single attach point at init.

What changed

C++ (src/common/worker/chip_worker.{h,cpp})

• ChipWorker::init now takes device_id and attaches the calling thread internally via simpler_init
• ChipWorker::set_device and ::reset_device deleted
• device_set_ flag and device_set() getter collapse into initialized_
• finalize() simplified — calls finalize_device_fn_ directly

C-API (pto_runtime_c_api.h + a2a3/a5 onboard+sim implementations)

• set_device export removed
• simpler_init signature extended to (ctx, device_id, log_level, log_info_v) and now returns int; folds the per-thread attach in alongside log config

DeviceRunner attach surface (a2a3 + a5, onboard + sim)

• attach_current_thread(int) is the centralized binder. Onboard: rtSetDevice. Sim: pto_cpu_sim_bind_device + pto_cpu_sim_acquire_device.
• device_id_ stays a plain int. All ChipWorker callers run on the same thread that called init(), so the standard thread-spawn happens-before edge is sufficient — no atomic, no CAS, no per-op re-attach.

Python

• _ChipWorker.init / ChipWorker.init take device_id (and the wrapper takes it before bins, since it's the simpler, more fundamental "where" parameter)
• _ChipWorker.set_device / .reset_device bindings removed; device_set property removed
• bootstrap_context no longer calls set_device — the device is attached at init time, the param is now used only as a defensive consistency check

Drive-by fix: ProfilerBase's constructor/destructor are now private with friend Derived (bugprone-crtp-constructor-accessibility). Newer clang-tidy versions on macOS would otherwise reject any TU that transitively includes profiler_base.h.

Callers updated

python/simpler/worker.py, simpler_setup/scene_test.py, tests/ut/py/test_chip_worker.py, tests/ut/py/test_worker/test_platform_comm.py, and test_bootstrap_context_{sim,hw}.py collapse init + set_device into a single init(device_id, bins) call.

Docs

docs/chip-level-arch.md, docs/dynamic-linking.md, docs/getting-started.md, plus the worker_malloc / hello_worker examples reflect the new flow.

Test plan

• `pip install --no-build-isolation -e .` — clean build (a2a3sim + a5sim) on macOS
• `pytest tests/ut/py` — 116 passed, 7 skipped (the only unrelated failures need `torch`, which is not installed in the local env)
• `python examples/workers/l2/{hello_worker,worker_malloc}/main.py -p {a2a3sim,a5sim} -d 0` — passes
• Onboard CI: `ut-a2a3` / `ut-a5` / `st-onboard-*` (Linux, hardware) — needs CI run
• `ctest --test-dir build/ut_cpp` (Linux) — needs CI run

[gemini-code-assist]

Code Review

This pull request refactors the ChipWorker API by integrating NPU device attachment into the init method and removing the explicit set_device and reset_device calls. Device operations such as malloc, free, and copy now internally handle thread attachment to ensure safety across different calling threads. Documentation, examples, and tests have been updated to reflect these changes. Feedback identifies potential data races in the initialization and attachment logic, along with a resource leak in the free_tensor error path.