Add AI Semantic Router Plugin

[danehans]

Add a plugin that utilizes Envoy External Processor (ExtProc) to act as an external Mixture-of-Models (MoM) router. The router should intelligently direct OpenAI API requests to the most suitable backend GenAI model from a defined pool based on the semantic understanding of the request's intent.

[danehans]

xref: https://github.com/redhat-et/semantic_router
xref: https://developers.redhat.com/articles/2025/05/20/llm-semantic-router-intelligent-request-routing

cc: @npolshakova @andy-fong @EItanya @DuncanDoyle

[danehans]

vLLM semantic router is a project that is gaining considerable interest in the k8s AI infra space. Similar to inference extension's EPP, the router is deployed as a separate component and uses ext-proc for proxy integration. Currently, the project's documentation and example k8s deployment guide use Envoy Gateway for the proxy component. Instead of kgtw natively supporting semantic routing, integrate with exteranl semantic routers and use vLLM semantic router as the first, supported implementation.

[danehans]

@linsun I set the priority based on our recent discussion. Please re-prioritize if needed.

[danehans]

xref vllm-project/semantic-router#382 for issues running the semantic router in k8s.