feat: add smart-llama - dynamic model loading layer for llama.cpp

[fdematos]

Summary

• Go server wrapping llama-server as subprocess for dynamic model loading
• Single active model with automatic swap when a different model is requested
• OpenAI-compatible API (/v1/chat/completions, /v1/models, /health)

Features

• Config: YAML-based configuration for models (native llama-server args format)
• Process Management: Subprocess lifecycle with health checks, graceful shutdown
• Reverse Proxy: Transparent proxying to internal llama-server
• CI/CD: GitHub Actions for tests/build on push/PR + multi-platform release on tags

Structure

cmd/smart-llama/       - Entry point
internal/config/       - YAML config loading
internal/process/      - Subprocess management
internal/proxy/        - HTTP reverse proxy
internal/server/       - HTTP server + routing
.github/workflows/     - CI + Release workflows

Testing

All packages include unit tests. Run with:

go test ./...