### The Feature

Native image generation routing for the EmpirioLabs provider, so this works without a manual `api_base`:

```python import litellm resp = litellm.image_generation(model=empiriolabs/qwen-image-2, prompt=...) ```

EmpirioLabs was recently added as a provider (#30278, docs in litellm-docs#341). Image generation already works today through the OpenAI-compatible route with an explicit `api_base` and `extra_body={sync: True}` (documented in litellm-docs#343), but the provider-prefixed form above does not resolve because images dispatch is driven by the static provider enum and routing code rather than the JSON provider registry.

### Motivation, pitch

We are happy to contribute this PR ourselves. Before writing it, we want to follow the architecture you prefer. From reading the code, supporting a JSON-registry provider in `image_generation` appears to touch the `LlmProviders` enum, the images routing in `main.py`/`images/`, and provider config resolution (5+ core files). Two possible shapes:

1. A conventional dedicated provider implementation for images (enum entry plus an images handler/transformation class, like existing image providers). 2. A more generic path where…