Draft: Fzconjecture optimization

The goal is to make the function FZ_conjecture_holds as fast as possible.

Possible avenues:

• optimize the multiplication of tautological classes to not go through prodtautclass and just write down the formula directly
• method 'pair': first intersect with monomials in kappa and psi-classes, before doing more complicated graphs
• make parallel implementation that runs intersections on multiple cores at the same time; here, the necessary depth of degeneration_graph should be run before each batch to have this as a common stored variable
• if FZ-conjecture is known in the relevant degree on some of the boundary divisors, can add the pullback-vectors to the matrix, and then exclude all generators that have an associated edge in their graph
• given some theoretical bound on the number of edges needed to represent any class in complementary degree, one could restrict to generators of at most that many edges; alternatively, explore which types of generators appear last in the list of basis elements