Apparent Psychological Profiles of Large Language Models are Largely a Measurement Artifact