Restarting from initial dump results in inital data ending up in the first dump

[pgrete]

@BenWibking discovered issues when restarting from an initial dump, see parthenon-hpc-lab/athenapk#120 for a reproducer.

The resulting output in dump 00001 contains data for t=0. All subsequent dumps are fine (i.e., 00002 contains the data that's supposed in 00002).
Also restarting from 00001 results in the outputs containing the correct data, so it's clearly a t=0 issue.