### What happened?

When a chat request to an `ollama_chat/*` model contains an `image_url` content block whose `url` is a **remote `http(s)://` URL**, LiteLLM forwards the URL string to Ollama's `images` field **verbatim** instead of fetching it and base64-encoding it. Ollama then tries to base64-decode the URL and fails:

``` litellm.APIConnectionError: Ollama_chatException - {"error":"illegal base64 data at input byte 5"} ```

Byte 5 is the `:` in `https:` — i.e. the raw URL is being treated as base64.

This contradicts the documented behavior in [`docs/proxy/image_handling`](https://docs.litellm.ai/docs/proxy/image_handling) ("Detect a URL being passed → Check if the LLM API supports a URL → Else, will download the base64 → Send the provider a base64 string"). Ollama does **not** accept URLs, so LiteLLM should be downloading + encoding here, but for the `ollama_chat` provider it does not.

### Relevant log output

``` litellm.APIConnectionError: Ollama_chatException - {"error":"illegal base64 data at input byte 5"} ```

### Steps to reproduce

```bash curl -s http://127.0.0.1:4000/v1/chat/completions \ -H "Authorization: Bearer $LITELLM_KEY" -H "Content-Type: application/json…