[Feature]: Custom TTS DSL support for initial websocket message

[rtATvw]

Problem statement

I'm trying to use vLLM Omni and it requires a session configuration message in it's setup.
https://docs.vllm.ai/projects/vllm-omni/en/latest/serving/speech_api/#streaming-text-input-websocket

Proposed solution

Add DSL a type to the request rules that allows sending a session configuration message at the start of the websocket connection.

Alternatives considered

Writing a proxy that sends the configuration message.

Impact area

None