Large Language Models are Capable of Offering Cognitive Reappraisal, if Guided