1) if you understand that the limits are x,z >= 0 and x+z<=2 then you either chose x:0->2 and hence z=2-x, or you choose z:0->2 and x=2-z. You are free to put it whatever way you want, with area integrals you choose one to be an independant range (the 0->2 is independant of any variables) and the other range must be dependant on this range.
2)He's cross-producted two vectors to obtain a normal vector....although i dont at this moment see where he grabbed those vectors from. It would be very easy to just look at the equation of the plane to see that the normal vector is just (2,1,2) since it's just the coefficients. Either way those two vectors must be vectors lying on the plane...so they might just have come from taking some basic points on the plane and subtracting one from the other to obtain a vector in the plane.
But then you raise an interesting question...i cant see why he's managed to leave the vector as a non-unit vector. Surely dS must be a unit vector?
Edit: I think i understand now: He is using a projection onto the xz plane....so the dxdz is smaller than the true area element ds...it must be multiplied up...and that has been encompassed in the use of (2,1,2) rather than the unit vector in this direction.