From Threat Report to ATT&CK: Automated Extraction and Reasoning of TTPs Using Large Language Models