Add support for GGUF models using wllama

[niutech]

It would be very useful to use GGUF models from Hugging Face using wllama, without converting to MLC or ONNX format. Please allow so.

[sauravpanda]

GGUF model support is now available via the Flare engine (PR #301, shipped in v2.1.0). Flare loads GGUF files directly in the browser via a Rust→WASM runtime with WebGPU acceleration.

const ai = new BrowserAI();
await ai.loadModel('llama-3.2-1b-flare');
const response = await ai.generateText('Hello!');

6 GGUF models are pre-registered. Custom GGUF URLs can be loaded via registerCustomModel(). This takes a different approach than wllama but solves the same user need. Closing — feel free to reopen if you specifically need wllama integration.

[niutech]

@sauravpanda Great, thank you!