Use of multiple imputation in supersampled nested case‐control and case‐cohort studies