## Describe the bug

When routing to `vertex_ai/claude-opus-4-7` (or `claude-sonnet-4-6`), LiteLLM unconditionally drops the `output_config` field from the request body before sending to Vertex's `streamRawPredict` endpoint. This silently disables the official Claude 4.7 **effort + task_budget** controls — the call succeeds, but `task_budget` is never honored and the model over-spends thinking on intermediate tool turns.

## Where

`litellm/llms/vertex_ai/vertex_ai_partner_models/anthropic/transformation.py` [lines 111-112](https://github.com/BerriAI/litellm/blob/main/litellm/llms/vertex_ai/vertex_ai_partner_models/anthropic/transformation.py#L111-L112):

```python # VertexAI doesn't support output_config parameter, remove it if present data.pop("output_config", None) ```

## Expected vs actual

Anthropic's "What's New in Claude 4.7" guide (https://platform.claude.com/docs/en/about-claude/models/whats-new-claude-4-7) documents the recommended invocation as:

```python client.beta.messages.create( model="claude-opus-4-7", max_tokens=128000, output_config={ "effort": "high", "task_budget": {"type": "tokens", "total": 128000}, }, betas=["task-budget…