Incorrect jaccard distance computed for features representing pairwise contacts

This is another subtask for treatment of contact features. The feature score model computes jaccard similarity metrics based on sets of labels present or absent on each sequence at each column of the alignment, and uses FeatureRenderModel.findFeaturesAtRes to compute the set of labels.

In v2.10 of Jalview and before, pairwise contact features are treated as interval ranges by this method, so the jaccard similarity will be different for three sequences where their disulphide bonded cysteines are properly aligned, but the sequence distance between cysteines in one protein is shorter or longer than the other.