Modified participation ratio approach: application to edge-localized states in carbon nanoclusters