vLLM 0.11 support

[erezzarum]

With the release of vLLM 0.11, the RunAI model streamer now breaks again.

vllm-project/vllm#26600

[noa-neria]

Upcoming version of vLLM will include an e2e test using public GCS bucket added here
We're also working on tests for our integration part in vLLM using mock object storage bucket