Simplify to single-model chat app: llama3.1:8b only. Add /api/chat SSE streaming, chatbot UI, remove model selection/download UI, build-time model pull

Fix Dockerfile: add zstd dependency for Ollama install

Dockerfile: auto-pull llama3.1:8b at startup in background. Syntax verified.

Replace simulated download with real ollama pull subprocess, add Ollama status check, installed models endpoint, WS wss support

Initial commit: FastAPI app with auth, API keys, download tracking, WebSocket live progress