RobustEncoder: Leveraging K-Means clustering technique to defend NLP models against backdoor attacks