uncertainty calcuation for time span, rather than specific time

Right now the uncertainty calculations (implemented in pydarm.uncertainty) takes a specific GPS time as input, with a "duration" specified in the uncertainty config. However, I think it would make more sense for the calculation to be over a time span (GPS start->stop), rather than just a specific time.

The primary motivation for this is the inclusion of the calibration uncertainty in the final aggregated GW frame files. The natural thing would be to include one set of uncertainty vectors that covers the entire time span of the frame. In principle we can know the start/stop time of the data in the frames where the uncertainty calculation will be included. We should therefore calculate the uncertainty based just on the data that is expected to be included in the frame file.

It's possible we could even have the uncertainty calculation be triggered by the frame aggregation process itself, with exactly the start/stop times to be included in the frame. That way we wouldn't have to worry about keeping things synchronized.