Investigating Large Language Models for Complex Word Identification in Multilingual and Multidomain Setups