Tree by sequence feature similarity doesn't exclude features in hidden columns

1. select columns 1-50 of the example uniprot alignment and deselect three sequences
2. Right click-> create features for the selcted region, and add a single Jalview feature.
3. Hit escape and then calculate->Calculate Tree-> Average distance based on sequence feature similarity
4. select columns 1-50 and hit 'H' to hide columns.
5. Calculate a second tree.

Result: Features trees are identical --> wrong tree produced when columns with features are hidden.

Cause:
- FeatureScoreModel.findDistances() operates on viewport.getAlignmentView().getVisibleAlignment()
- this is an alignment with just the visible sequences, in contiguous columns
- method FeatureRenderer.findFeaturesAtRes(seq, pos) is used to find features
- this method doesn't adjust for hidden columns
- e.g. if columns 11-20 are hidden, it will find features in columns 11-20 as if for columns 21-30