A machine learning model incorporating 18F-prostate-specific membrane antigen-1007 positron emission tomography/computed tomography and multiparametric magnetic resonance imaging for predicting prostate-specific antigen persistence in patients with prostate cancer after radical prostatectomy