Add Qwen3.5 and a broader export path

### Source link

https://huggingface.co/collections/Qwen/qwen35

### Target platform

iOS

### Use case

Could you add Qwen3.5 (up to 4B) for on-device use? A broader Hugging Face → .aimodel export path would also be valuable, since model-by-model support does not seem sustainable as new architectures keep appearing. Thank you in advance!

### Preferred precision / compression

No preference. A 4-bit option would be useful for on-device testing, with FP16 available for comparison.

### Additional context

_No response_