[docs] [llm] [template] update ray version + use anyscale image for the gpt-oss deployment

[Aydin-ab]

Updating to a more recent release of ray where ray serve llm is more stable with vllm. Also using an anyscale image instead of a dockerfile in the anyscale service config file for gpt-oss. Previously, we used a dockerfile because there was compatibility issues with vllm but this has been fixed. Building from a dockerfile is a pretty bad experience for a user (can take ~4-5 min for the service to spin up)

[gemini-code-assist]

Code Review

This pull request updates the Ray and vLLM versions in several documentation tutorials and switches the gpt-oss tutorial to use a pre-built Anyscale image instead of a Dockerfile, which is a great improvement for user experience.

I've found a few inconsistencies in the gpt-oss tutorial documentation (README.md and notebook.ipynb) where some values seem to have been copied from another tutorial. I've left specific comments with suggestions to fix them.

Additionally, I noticed that in the "Custom Dockerfile" examples across several of the updated markdown and notebook files, the commented-out image_uri still refers to the old version 2.49.0. While these are not part of the diff, it would be good to update them to 2.52.0 for consistency.

Overall, the changes look good and improve the documentation.

[kunling-anyscale]

looks good, left a few suggestions