#32216 [Feature]: Auto-route Vertex AI Interactions passthrough by project and location
### Check for existing issues
- [x] I have searched the existing issues and checked that my issue is not a duplicate.
### The Feature
On the Vertex AI pass-through route, add opt-in auto-routing for the Interactions API so that a create call resolves the GCP project and location from the request body's model, and the follow-up get/delete/cancel calls route back to the originating project without the caller having to track it. Gate it behind a `general_settings` flag, default off
### Motivation, pitch
A `generateContent` pass-through request already auto-corrects the project and location from the model id in the URL, so a caller can send a placeholder project and still reach the right backend. The Interactions API cannot use that path: on create the model is in the request body, and get/delete/cancel carry neither a model nor a body. Because the pass-through is stateless and selects credentials purely from the URL's `{project}-{location}`, an async interaction (for example a background Omni Flash video job) created under one project fails to poll unless the caller reconstructs the exact same project in every follow-up URL, which is easy to get wrong across multiple projects. En…