The `heuristics` function takes as input a distance matrix and returns prior indicators of how bad it is to include each edge in a solution. The return is of the same shape as the input. The heuristics should contain negative values for undesirable edges and positive values for promising ones.