Stack based refactor pt. 2

[nossleinad]

Translate branchlength_optim! and nni_optim! from recursive to iterative

• Use Iterators.reverse to avoid allocs and enforce a more consistent style in stack-based implementations
• Don't ignore starting_index in set_node_indices

Note:
We need to decide if we want to go with nni_optim! or what I call nni_optim_full_traversal!. The latter ensures that all nodes are visisted during the optimization which the former does not.

[murrellb]

We need to decide if we want to go with nni_optim! or what I call nni_optim_full_traversal!. The latter ensures that all nodes are visisted during the optimization which the former does not.

Do you have any side-by-side comparisons for tree search using both options? Try starting from random, vs starting from a nearby tree, etc?

[nossleinad]

Starting from a random tree

Starting from a nearby tree

[nossleinad]

Full tree_polish! starting from a random tree

Full tree_polish! starting from a nearby tree