Update llama.cpp to support Qwen 3.5 architecture

[bereilhp]

Issue

The currently pinned llama.cpp commit does not recognize the qwen35 model architecture.

Error when loading a Qwen 3.5 GGUF model:

unknown model architecture: 'qwen35'

Request

Update the llama.cpp submodule to a more recent commit that includes support for the qwen35 architecture.

References

Support was added in commit fc0fe4004985d6749a7a05e250d161f9dbe41d65, introduced in this PR: ggml-org/llama.cpp#19468