v0.8.0

Improve Qwen-style thinking-template cleanup and update llama.cpp to b9009.

SwiftPM binary artifact for CarbocationLocalLLM's llama runtime.

This archive redistributes static llama.cpp/ggml object code through
libllama-combined.a. See THIRD_PARTY_NOTICES.md in this release tag
for bundled and linked notices.

Checksum:
57a658f427fc3f900d6f9499d59f6accbcfb482dae933514599d96ed1b3d614b