Remove Langfuse Dependency for Session ID Grouping in Response API

Currently, in the Response API, we derive the session_id from Langfuse by querying it using the previous response_id, as shown below:

if response_id:
    traces = self.langfuse.fetch_traces(tags=response_id).data
    if traces:
        self.session_id = traces[0].session_id

This approach introduces an unnecessary runtime dependency on Langfuse, solely to group related traces under the same session for visualization purposes.

OpenAI does not currently provide a native conversation ID for API responses, which complicates session tracking.
We are in discussions with OpenAI to potentially add support for a conversation_id or similar identifier at their end which we can pass to langfuse as session_id.

Alternative: Maintain the session or conversation-level grouping within the Glific flow context itself, so that Langfuse becomes an observer rather than a source of truth for session identity.

El issue identifica la ruta de Response API y la búsqueda existente de trazas de Langfuse como punto de entrada; no se menciona ningún archivo ni prueba. Primero localiza dónde se gestionan response_id y session_id y, después, verifica que la agrupación de sesiones siga siendo coherente sin una consulta de Langfuse en tiempo de ejecución.