Auto-resume on startup; fix Groq sleep bug and env var model selection

Add Groq LLM provider and fix speed controls for real fast-forward

Scale to 100 agents: procedural personas, daily routines, speed controls

Switch to llama3.1:8b with Ollama native JSON mode

Fix Windows encoding issues: UTF-8 log file and ASCII arrow in logging

Add Ollama support for free local LLM inference

Initial implementation of Soci city population simulator