FM Phase 6: Context-window auto-detection #27

On startup / first /chat, probe the inference endpoint for context size.

• Ollama /api/show, llama.cpp /props, vLLM /v1/models
• Cache in config / in-memory, log the detected value
• Ship known-model lookup table as fallback
• Unknown → conservative defaults, user can override
• [inference] max_context in config is authoritative when set

Files: inference.cpp, config.cpp