Turing, Inexacterror for DiscreteUniform distribution, with NUTS sampler

You can build a GibbsSampler where you sample alpha with MH and the rest with NUTS