[Doc]: Update the vllm distributed Inference and Serving with the new MultiprocessingGPUExecutor

[rcarrata]

📚 The doc issue

The vLLM documentation only reflects the possibility to use Ray for running Distributed Inference and Serving with vLLM, even though the #4539 issue is merged and v0.4.3 is released with the MultiprocessingGPUExecutor feature included as an alternative to Ray for single-node inferencing.

Suggest a potential alternative/fix

Update the documentation to reflect the possibility of using MultiprocessingGPUExecutor as an alternative to Ray for single-node inferencing.

[youkaichao]

Contributions are welcome!

[njhill]

I have included these doc updates in #5230.

[rcarrata]

thanks for your help @njhill, much appreciated!