Task durations about cluster-data HOT 2 CLOSED

The dataset is a time-series, so there's actually no guarantee that the entries 
for each 
task are the first we're seeing of them, nor the last, and there might be tasks 
that 
started before the dataset that end in the duration of this dataset, or tasks 
that end 
after the duration of this dataset, so I think in the current form, it's not 
really feasible to 
have task durations. Maybe in another iteration, future datasets could include 
only 
complete tasks? But that might lead to anonymity problems so I don't know.

The new format ([TraceVersion2]) will address this issue.