Advancing Air Quality Monitoring: TinyML-Based Real-Time Ozone Prediction with Cost-Effective Edge Devices