Training Heterogeneous Client Models using Knowledge Distillation in Serverless Federated Learning