Add a function to customize deletion order for shortcut testing

Simply deleting hypotheses in a custom order is not difficult. The main hurdle is figuring out valid deletion orders. First, all rejected hypotheses must come before non-rejected, which is easy once you have initial test results. But the rejected ones cannot necessarily go in any order, as deleting one hypothesis increases weights of other hypotheses. The simplest route from a coding perspective would probably be to just loop over every permutation of rejected hypotheses and mark them as valid or not. But this would get slow very quickly. I think there's some way to do it using tricks with the closure. Like checking each row of the closure to see if there's a path to it using its parent rows. I guess parents of each row could be calculated on the fly by replacing each 0 with a 1, that might not be so slow. But there should be a formula method similar to what's used in graph_generate_weights().