What is the purpose of the skip_end argument in learn.recorder.plot function as shown in the NLP, Tabular Data lesson?