[Doc] Improve quickstart documentation

[rafvasq]

This PR tries to improve the existing quickstart guide for newcomers to the project who want to generate something with vLLM quickly.

• Include pre-req system requirements
• Include installation step
• Reorganize sections
• Add links to help with LLM terminology
• Improve step-by-step guide readability
• Use Qwen2.5-1.5B-Instruct for chat/server sections

[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.

🚀

[ckadner]

Thank you @rafvasq for improving the quickstart guide. I have a few nits and some questions. I'm new rst formatting as and some of my suggestion are likely just personal preference. Feel free to ignore those :-)

[rafvasq]

Thanks @ckadner, I believe I addressed everything!

[ckadner]

Looks good. Thanks @rafvasq

[simon-mo]

This looks great! Can you also update the chat/server section to ues https://huggingface.co/Qwen/Qwen2.5-1.5B-Instruct so that the users don't need a chat template to start with (advanced feature)?

[rafvasq]

! Can you also update the chat/server section to ues https://huggingface.co/Qwen/Qwen2.5-1.5B-Instruct so that the users don't need a chat template to start with (advanced feature)?

Hey @simon-mo, would removing the extra note about chat templates suffice? It doesn't seem to me like a user needs to know anything about chat templates to use the guide with facebook/opt-125m as written since vLLM just uses a default quietly.

[simon-mo]

ah that's a good point. however my main problem with the opt model is that just generates gibberish for chat.

[rafvasq]

Makes sense, I've updated the model for those sections