Large Legal Fictions: Profiling Legal Hallucinations in Large Language Models