## Bug Description

`litellm.get_supported_openai_params(model='gemini/gemini-2.5-flash')` includes `frequency_penalty` in the returned list, but when the parameter is actually passed, the Gemini API rejects it with a 400 error:

``` Penalty is not enabled for models/gemini-2.5-flash ```

This also means `litellm.drop_params = True` does not help — since litellm considers the param "supported" for Gemini, it maps it to the native Gemini API parameter rather than dropping it. The API then rejects it.

## Reproduction

```python import litellm

# This claims frequency_penalty is supported: params = litellm.get_supported_openai_params(model='gemini/gemini-2.5-flash') print('frequency_penalty' in params) # True

# But this fails: litellm.drop_params = True # has no effect since param is "supported" response = await litellm.acompletion( model="gemini/gemini-2.5-flash", messages=[{"role": "user", "content": "Hello"}], temperature=0, max_tokens=50, frequency_penalty=0.4, ) # GeminiException BadRequestError - {"error": {"code": 400, "message": "Penalty is not enabled for models/gemini-2.5-flash"}} ```

## Expected Behavior

Either: 1. `get_supported_openai_params` sho…