The use of islands and cluster-randomized trials to investigate vector control interventions: a case study on the Bijagós archipelago, Guinea-Bissau