Pypi build for aarch64

[surak]

Currently, the python wheels are for x86_64 only. It would be good to have them on arm as well.

[j93hahn]

@surak follow the instructions to build from source, it worked very smoothly for me

[surak]

@j93hahn Oh, I know how to build from source :-) But I don't want to do that for every version, at every supercomputer we have here with a different architecture :-)

[drikster80]

+1 for releasing an official aarch64 whl file in addition to the x86_64. This would help downstream projects such as VLLM provide official docker (vllm-project/vllm#10499) without the need to continuously rebuild and support a from-source build.

[Skylion007]

+1 I'm running into issues deploying this on an HPC cluster due to the difficulty with the getting the arm64 version compiled. It's the main blocker from me using GH200 style clusters.