Add vllm v1 support for LocalVLLMBackend

[psschwei]

vLLM v0 has been deprecated, but currently LocalVVLMBackend does not support v1. As a result, vllm appears to be pinned to release 0.9.1 (from June 2025).

I'd propose adding VLLM v1 support for the LocalVLLMBackend. I believe that should also allow for Remote VLLM (via OpenAI API) to work, but will need to verify that.

[psschwei]

xref #288