 Gemma 3n 4-bit finetune Q4_K_M GGUF support

Q4_K_M is not supported:

```py
RuntimeError: Unsloth: `q4_k_m` quantization type is not supported.
The following quantization types are supported: `['f32', 'f16', 'bf16', 'q8_0']`
```

_Originally posted by @aphilas in https://github.com/unslothai/unsloth/discussions/2924_