DIViS: Domain Invariant Visual Servoing for Collision-Free Goal Reaching