[Bugfix][CI/Build] Fix docker build where CUDA archs < 7.0 are being detected

[LucasWilkinson]

FIX #9226

In this dockerfile the archs being detected by torch is 5.0;8.0;8.6;8.9;9.0;9.0a we should ensure we prune off 5.0 for the kernels where we build for all target archs. This PR does that by pre-filtering the CUDA_ARCHS (target archs) by the CUDA_SUPPORTED_ARCHS:

  cuda_archs_loose_intersection(CUDA_ARCHS 
    "${CUDA_SUPPORTED_ARCHS}" "${CUDA_ARCHS}")

(PR also includes some logging improvements / comments)

Verified using:

FROM pytorch/pytorch:2.4.0-cuda12.1-cudnn9-devel AS build

ARG GITHASH="e232f585359d50e880bcc4aa8c4161ab17919567"

RUN apt update && apt install gcc g++ git -y && apt clean && rm -rf /var/lib/apt/lists/*

ENV PATH=/workspace-lib:/workspace-lib/bin:$PATH
ENV PYTHONUSERBASE=/workspace-lib

RUN pip install git+https://github.com/neuralmagic/vllm.git@$GITHASH --no-cache-dir --user -v

FROM pytorch/pytorch:2.4.0-cuda12.1-cudnn9-runtime AS vllm-openai

WORKDIR /vllm-workspace

COPY --from=build /workspace-lib /workspace-lib

ENV PATH=/workspace-lib:/workspace-lib/bin:$PATH
ENV PYTHONUSERBASE=/workspace-lib
ENV PYTHONPATH=/workspace-lib:/vllm-workspace

ENTRYPOINT ["python3", "-m", "vllm.entrypoints.openai.api_server"]

[github-actions]

👋 Hi! Thank you for contributing to the vLLM project.
Just a reminder: PRs would not trigger full CI run by default. Instead, it would only run fastcheck CI which starts running only a small and essential subset of CI tests to quickly catch errors. You can run other CI tests on top of those by going to your fastcheck build on Buildkite UI (linked in the PR checks section) and unblock them. If you do not have permission to unblock, ping simon-mo or khluu to add you in our Buildkite org.

Once the PR is approved and ready to go, your PR reviewer(s) can run CI to test the changes comprehensively before merging.

To run CI, PR reviewers can do one of these:

• Add ready label to the PR
• Enable auto-merge.

🚀

[mgoin]

This is good by me

[tlrmchlsmth]

LGTM, thanks for the fix!

I saw one typo in a comment but I think we just land this to get the fix in.

[tlrmchlsmth]

Suggested change

	#
	# For cuda we want to be able to control which architectures we compile for on
	# a per-file basis in order to cut down on compile time. So here we extract
	# the set of architectures we want to compile for and remove the from the
	# CMAKE_CUDA_FLAGS so that they are not applied globally.
	#
	#
	# For cuda we want to be able to control which architectures we compile for on
	# a per-file basis in order to cut down on compile time. So here we extract
	# the set of architectures we want to compile for and remove them from the
	# CMAKE_CUDA_FLAGS so that they are not applied globally.
	#