This makes post-processing handle the case that the departure time only
contains a time by using the departure day value to fill the gap.

This avoids extractor scripts having to do error prone date/time
computations in JS, in particular if both values are extracted from
different document nodes (e.g. a barcode and a PDF body).