[bugfix] actually adjust trace_start_time in channelStopFilter

[sjoerd-bouma]

While going through the definitions of times in #773, I noticed that the channelStopFilter doesn't actually adjust the trace_start_time. This is a very minor bug, as all traces are affected the same way, but it's probably worth doing it correctly.

[fschlueter]

I think you have to change the trace_start_time by int(np.round(prepend * sampling_rate))) / sampling_rate. And you can use add_trace_start_time(...)` instead

[fschlueter]

I think you have to change the trace_start_time by int(np.round(prepend * sampling_rate))) / sampling_rate. And you can use add_trace_start_time(...)` instead

or simply use prepend which should be already in units of time but not prepend * sampling_rate

[fschlueter]

And I guess you have to update the reference values for the failed reconstruction?

[sjoerd-bouma]

Thanks Felix, I made a mess of that one... should be correct now, I hope. I've only adjusted the signal_time in the reference.json for now, just to demonstrate that this is the only difference; however, I noticed that the tiny_reconstruction test complains about a lot of missing keys because we've added some more stationparameters since it was originally constructed (see the test output on develop, for example).

I would suggest I just generate a completely new reference file that includes all current parameters to get rid of all these warnings before merging this (and maybe reformat the json so git has an easier time comparing changes in the future...)