feature: Native support for v1/responses endpoint

Did you check the docs?

• I have read all the NeMo-Guardrails docs

Is your feature request related to a problem? Please describe.

OpenAI's v1/responses endpoint is the de-facto API for agentic tool-use, superseding v1/chat/completions. While OpenAI still supports the Chat Completions API, the Responses API is far more performant agentic use cases. Furthermore, OpenAI recommends new projects to use Responses over Chat Completions.

Currently, there is no native support for v1/responses payloads in NeMoGuardrails. Developers who rely on the Responses API for their applications would have to write custom code to translate v1/responses payloads to v1/chat/completions in order to use NeMoGuardrails.

Describe the solution you'd like

• A native /v1/responses client, mirroring the existing /v1/chat/completions client
• A /v1/responses endpoint in the NeMo Guardrails server, for guardrailed inference proxying
• A /v1/responses/checks endpoint, for standalone guardrailing

Describe alternatives you've considered

n/a

Additional context

No response

Start by locating the existing /v1/chat/completions client and server endpoint, then trace how their payloads and standalone checks are handled. Done means native /v1/responses and /v1/responses/checks support is available alongside the existing client and endpoint behavior.