### Check for existing issues

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

### The Feature

A supported way to make LiteLLM send Vertex AI's **native `responseSchema`** (OpenAPI-style) instead of **`responseJsonSchema`** for Gemini 2.x+ models when a `json_schema` `response_format` is passed. Either of these shapes would work:

- a per-request param, e.g. `litellm.completion(..., vertex_schema_format="response_schema" | "response_json_schema")`, or - a module-level setting, e.g. `litellm.vertex_ai_use_response_json_schema = False`

Today the choice is hardcoded and cannot be influenced by any config, env var, or request param (verified against v1.93.0 source). The fork lives in `VertexGeminiConfig.apply_response_schema_transformation` (`litellm/llms/vertex_ai/gemini/vertex_and_google_ai_studio_gemini.py`):

```python use_json_schema = supports_response_json_schema(model) # regex on the model NAME only ... if use_json_schema: optional_params["response_json_schema"] = _build_json_schema(deepcopy(schema)) # verbatim passthrough else: optional_params["response_schema"] = self._map_response_schema(value=schema) # native conversi…