A Hot-Deck Multiple Imputation Procedure for Gaps in Longitudinal Data on Recurrent Events