[EP] Support vLLM; high-throughput and low-latency

[MaoZiming]

Description

Please include a summary of the changes and the related issue.

Fixes # (issue)

Type of Change

• Bug fix
• New feature
• Documentation update

How Has This Been Tested?

Include any tests here.

• Unit tests
• Integration tests
• Manual testing

Checklist

• My code follows the style guidelines, e.g. format.sh.
• I have run build_and_install.sh to verify compilation.
• I have removed redundant variables and comments.
• I have updated the documentation.
• I have added tests.

[MaoZiming]

bash launch_vllm_head.sh 10.4.147.22 13345 deepseek-ai/DeepSeek-V3-0324 deepep_low_latency 2 1 8 1
bash launch_vllm_worker.sh 10.4.147.22 13345 deepseek-ai/DeepSeek-V3-0324 deepep_low_latency 2 1 8 1

This works.

[MaoZiming]

Using this commit b1029ad right now, there seems to be some changes from merging master. I need to look more.

[MaoZiming]

#620
This seems to break the vLLM high-throughput mode integration. Marking it here to double check later.
I got

[wait_until_cmd_consumed nvl:5 cmd:4 label:3] waiting slot=565