Add qwen3-embedding-0.6b-16k embedder llamafile (from claude-hooks v1.4.0, sha 414f6166) for opencoti-server download+install testing (F2 M8 / F4 M8)

release layout: archive c3 under releases/archived/ (keep only latest cut at root)

c4: sbsa side-load DSO

c4: x86_64 side-load DSO

c4: aarch64 artifact

c4: win variant

c4: x86_64 artifact

Upload dso/0.10.3-c3/ggml-cuda-sbsa-aarch64.so with huggingface_hub

Upload opencoti-llamafile-0.10.3-c3-aarch64.llamafile with huggingface_hub

c3: standalone side-load CUDA DSO (x86_64-linux, for the -win/bare variant)

c3: add -win variant (bare thin APE, <4GB Windows limit)

Upload opencoti-llamafile-0.10.3-c3-x86_64.llamafile with huggingface_hub

initial commit