Excited-state nonadiabatic dynamics in explicit solvent using machine learned interatomic potentials