[zhangzhenghao]

Summary

• Use std::thread::hardware_concurrency() for automatic CPU inference threads, guarded to at least one thread and without the previous eight-thread cap.
• Preserve explicit positive thread overrides and existing Python/Node/CLI API capabilities.
• Python/Node tests and examples, and documented CLI examples, omit explicit thread arguments.
• Delete the requested README paragraph without replacement.
• Remove tests/cpu-threads-runtime-test.cpp and its CMake registration as requested.

Verification

• Latest update: reconfigured and rebuilt the Linux Release static build; CTest 33/33 passed.
• Earlier revision: shared CTest 32/32 passed, actual published-model Python/Node tests passed, and all nine remote CI checks passed.
• CI for the latest test-removal commit is pending; earlier CI results do not verify the new head.