Trying Latin Hypercube sampling in generating input ASTs

[galaxyumi]

I just learned that there is a good method we can try to generate input ASTs. The method is called 'Latin Hypercube sampling' (https://en.wikipedia.org/wiki/Latin_hypercube_sampling). I haven't looked into any details yet, but I think we definitely want to have some discussion about how it works, how efficient this method is, how much improvement we can achieve in evaluating observational models, etc. in the future!