On COVID-19 Prediction Using Asynchronous Federated Learning-Based Agile Radiograph Screening Booths