On the Generalization of Sleep Apnea Detection Methods Based on Heart Rate Variability and Machine Learning