Gravar-mail: Handling misclassified stratification variables in the analysis of randomised trials with continuous outcomes