Speaker Fuzzy Fingerprints: Benchmarking Text-Based Identification in Multiparty Dialogues