Generates spatially coherent 3D noise based on sparse convolution.
sturb(X, Y, Z, depth)
Spatially coherent noise gives random numbers which are close to each other when the X, Y, Z inputs are close to each other.
depth is the amount of fractalization of the noise.
This is very similar to turb, except instead of finding noise values on a fixed lattice and interpolating, this function finds noise values on points scattered in space and interpolated according to the Voronoi decomposition.