[Workers AI] Undefined result for models that return binary output

### Confirm this is a Typescript library issue and not an underlying Cloudflare API issue

- [x] This is an issue with the Typescript library

### Describe the bug

Hello!

I can't tell if I'm holding the SDK wrong, but when I use the Typescript SDK to run a model like [`leonardo/phoenix-1.0`](https://developers.cloudflare.com/workers-ai/models/phoenix-1.0/) that, according to the docs, returns the following payload...

```json
{
"type": "string",
"contentType": "image/jpeg",
"format": "binary",
"description": "The generated image in JPEG format"
}
```

...I get an undefined output. If I run a model that returns a base64 string (like [`schnell`](https://developers.cloudflare.com/workers-ai/models/flux-1-schnell/)), I get back a JSON object that correctly reflects the type that's in the documentation.

What could I be doing wrong?

### To Reproduce

Run this snippet:

```ts
import Cloudflare from 'cloudflare';

const client = new Cloudflare({
apiToken: process.env['CLOUDFLARE_API_TOKEN'], // This is the default and can be omitted
});

const output = await client.ai.run("@cf/leonardo/phoenix-1.0", {
prompt: "A cat eating soup dumplings",
account_id: process.env['CLOUDFLARE_ACCOUNT_ID']!,
})
console.log('Output type:', typeof output);
console.log('Output:', output);
```

Observe this output:
```
Output type: undefined
Output: undefined
```

### Code snippets

```Typescript

```

### OS

macOS

### Runtime version

Typescript 5.7.3

### Library version

v5.2.0

Start at the client.ai.run entry point shown in the reproduction and trace how responses are handled for the binary-output model versus the base64-string model. Run the provided snippet against leonardo/phoenix-1.0 and confirm that the binary result is returned instead of undefined; add coverage for this behavior if the surrounding code has tests.

Written by the indexing model from the issue text.