Don't attempt concrete eval if there's no information to be gained (#46774)
If the effects are already maximized and the rt is already Const,
there is nothing concrete evaluation could possibly tell us
that we don't already know - just bail early in that case to
save some inference time.