Trace gap tolerance - allow for linear interpolation?

[tmgreenfield1101]

Currently a station is removed if it has a gap of more than 1 sample. This is silly as linearly interpolating this would produce a signal that is fine to process. Instead we should define a minimum gap before QM removes the station. this could be user defined - but most users are not likely to come across it. maybe a default of 10 or 100 samples is appropriate.

[hemmelig]

A sensible parameter name might be something like 'gap_tolerance' or 'time_tolerance' (as per http://ds.iris.edu/SeismiQuery/timeseries.htm - "Time Tolerance defines the tolerance at which to consider a time difference to be a gap/overlap or continuous. For instance, if a time tolerance of 1 second is specified, then any time tears or overlaps of 1 second or less will be ignored and that data will be considered continuous.") I think 'gap_tolerance'.

The choice of units is either seconds or number of samples. Will need to consider what we want to be consistent between data with differing sampling rates. Using units of sample makes setting a default simpler across multiple sampling rates. The ObsPy method we use to look for gaps is referenced in seconds, but we have the data sampling_rate to hand to convert between samples and seconds.

This would be added to the QMigrate.io.data.Archive class. I agree some tolerance would be useful, but still think a default of 1 sample is best. Extremely gappy data might be indicative of some underlying problem with the data that would make it unsuited for QuakeMigrate.