Pre-Training Language Models for Identifying Patronizing and Condescending Language: An Analysis