Wait... I think I got it..
Pick any y' and x'. Then, g(y') ≤ h(x,y') for all x; therefore g(y')≤h(x',y'). Now, h(x',y')≤sup{h(x',y) : y in Y}=f(x'). This established the following inequality:
g(y')≤f(x').
Because x' and y' were arbitrary, we conclude g(y)≤f(x) for all x,y. Thus g(y) is...