Batched Learning in Generalized Linear Contextual Bandits With General Decision Sets