Update llama.cpp for Qwen 3.5 support

[bereilhp]

Background

The currently pinned llama.cpp commit does not recognize the qwen35 model architecture. This causes the application to fail when loading Qwen 3.5 GGUF models with the error: unknown model architecture: 'qwen35'

Summary

Updated the llama.cpp dependency in package.json to commit fc0fe4004985d6749a7a05e250d161f9dbe41d65. This version includes the necessary implementation to support the qwen35 architecture.

Manual Verification

1. Ran pnpm clean to remove previous build artifacts.
2. Ran pnpm install to fetch the updated llama.cpp source commit and recompile the native bindings.
3. Ran pnpm build:ts to compile the project's TypeScript source.
4. Executed pnpm run test:e2e using the Qwen 3.5 model (Qwen3.5-4B-Q4_K_S.gguf) to verify end-to-end functionality.
5. Verified that the "unknown model architecture" error was resolved and the model loaded correctly.

Future Work

None.

Related Issues

Fixes #26