Analyzing patient‐reported outcome data when completion differs between arms in open‐label trials: an application of principal stratification