Add support for Qwen3-vl models

**Is your feature request related to a problem? Please describe.**
When trying to load Qwen3-vl models like Qwen3-VL-235B-A22B-Thinking or Qwen3VL-8B-Instruct, get error messages as following:
site-packages/llama_cpp/_internals.py", line 78, in close
if self.sampler is not None:
AttributeError: 'LlamaModel' object has no attribute 'sampler'

**Describe the solution you'd like**
Looks that Qwen3-vl models are not supported yet, could you help add the supporting for these Qwen3-vl models?