[Installation]: Compiling using `build_wheel.py` helper script uses logical instead of pysical cores for compilation

System Info

System Information:
I run inside the latest version of the TensorRT-LLM development container

How you are installing TensorRT-LLM

Using ./scripts/build_wheel.py --clean --cuda_architectures "89".

I got OOM compiling, and while debugging the build_wheel.py code, I noticed that by default, parallelism is set to multiprocessing.cpu_count(), which is the number of logical cores. However, compilation process is CPU intensive. Ideal number of thread for compilation jobs +- number of physical cores.
The current default increases memory usage, and costs a lot of context switches, with little increased parallelism.
Therefore in scripts/build_wheel.py, the --job_count should default to number of physical CPUs.

Before submitting a new issue...

• Make sure you already searched for relevant issues, and checked the installation documentation and examples for answers to frequently asked questions.