Monte Carlo integration: why the error falls as one over the square root of N

Afterwards you can estimate an integral by random sampling, give it an error bar, and say why its error falls as one over the square root of N in any dimension.