Performance optimisation in fetch_all_by_feature_class method. Contributed by Shiran Pasternak at CSHL

load different Go sources as the same so that the xrefs are NOT treated as priority ones which only give one to one mappings, which is in approriate for GO

  The ranges in the registry are sorted by start position.  When
  calculating the overlap with a query range, a loop goes through the
  registered ranges and checks for overlaps between each range and the
  query.  This change will terminate this loop when all the remaining
  start positions are larger than the end position of the query range.

  This cuts the running time for this routine by 90% in some
  applications that registers a large number of ranges.

Tested.