Development and verification of a combined diagnostic model for primary Sjögren’s syndrome by integrated bioinformatics analysis and machine learning