Determinants of the energy costs of light activities: inferences for interpreting doubly labeled water data