Wrong dtype for Geant4 time field.

[HenningSE]

With #195 we changed the dtype for the Geant4 input "t" from float64 to int64. This was discussed here: #195 (comment). In fact I now realized that this change is wrong and can lead to problems with the upcoming lineage clustering where the exact timing of our energy deposits is of interest. To simulations running with DBSCAN clustering, the effect should be minimal.