### Check for existing issues

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

### What happened?

When using a model configured with `model_info: { mode: audio_transcription }` using the `hosted_vllm/` provider, the `ssl_verify` parameter specified in the `litellm_params` is ignored by the transport layer.

While standard chat completion and embedding requests respect `ssl_verify`, requests routed through the `audio_transcription` mode fail with SSL verification errors when the backend uses custom CA certificates or self-signed certificates.

**Expected Behavior:** The `ssl_verify` path should be used to configure the underlying HTTP client (e.g., `aiohttp` or `httpx`) to trust the provided CA certificate.

**Actual Behavior:** The `ssl_verify` parameter is ignored by the network client and is instead incorrectly treated as a request payload parameter, leaking into the `extra_body` of the API request.

### Steps to Reproduce

1. Set up a `config.yaml` for LiteLLM Proxy with a model using the transcription mode and a custom SSL certificate: ```yaml model_list: - model_name: audio-model-placeholder model_info: mode: audio_transcr…