Consider using updated NumPy RNG

[chimaerase]

Thanks for your work on this library! I was looking over the Latin Hypercube Sampling and noticed that it uses the legacy NumPy RandomState to generate random numbers instead of numpy.random.Generator that replaces it (documentation mentions this in the first sentence). I realize a change might break repeatability in some legacy code, though I think there's also an argument for encouraging use of modern NumPy API's that will get continuing improvements.