The third methode is wrong because of the input RC delay and integrative configuration.
The first and the second methode does not measure the real slew rate of the fully differential output, the slew rate will be different for inverting input (image 2) versus noninverting input configuration (image 1).
However any of the measuremet schemes 1 or 2 will gave you an approximative value of the slew rate. If you need a precise one, use a fully differential generator and measure the differential output value with a dual channel scope.