I think that’s because in computer science most master/slave nomenclature comes from hardware with a command/control structure (still notable in things like Spark where the namenode/master node controls the data nodes).

GIT just took naming conventions from other existing design patterns (although I should probably look up sources to verify that assumption).