Fix documentation on Qwen3-VL usage

[kovtcharov]

Currently, Qwen3-VL is being run using iGPU via llama.cpp/Vulkan, the documentation incorrectly asserts NPU execution for the EMR agent example.