Add a lower bound for repro error threshold

[J.D. Laurence-Chasen]

In many cases, an accurately tracked point has a non-zero reprojection error average. In these instances, a 'better' reprojection error for a given frame may actually indicate a bad track. It would be nice to be able to set a lower, second bound for reprojection error, so when I click '4' to progress through frames that exceed the threshold, instances where the error deviates from the overall average in both directions are included.