Ollama vs. vLLM: When to Start Scaling Your Local AI Stack