Compute probabilities sequentially

[Leguark]

Make a function that the probabilites arrays (one per lith) are created at the same time we are making the inference. This is useful to avoid huge memory consumption when many iterations with high resolution are used