### Check for existing issues

- [x] I have searched the existing issues and checked that my issue is not a duplicate.

### What happened?

Creating a Vertex AI batch with `vertex_location: global` is **accepted** at request time (no validation error), but litellm then builds a request URL using that location and forwards it to Vertex, which responds with an **HTTP 400** downstream. So `global` is silently accepted as if it were a valid Vertex region for batch operations, and the failure only surfaces well after the request has already been accepted — instead of failing fast with an actionable error, or being handled correctly.

**Best-effort root cause (please confirm/correct):** litellm's Vertex request-URL construction (which builds the per-region `*-aiplatform.googleapis.com` endpoint from `vertex_location`) doesn't special-case `global` for the batch surface the way some other Vertex AI endpoints do, so it builds a URL Vertex's batch API rejects.

### Steps to Reproduce

1. Configure a Vertex AI model/`litellm_params` with `vertex_location: global`. 2. Create a batch via `POST /v1/batches` (or the SDK equivalent) against that model. 3. Observe the create call fails with an HTT…