This PR fixes the problem of miscounts when the tree is a DAG instead of a strict tree, by recursively building up the list of types of nodes as a hash and returning the node ids instead of the counts. That way they don't get double counted.
Also, adds history to time (teal) nodes and calculates a linear regression on that history to make a prediction about when the number of incomplete smalls goes to 0.
This PR fixes the problem of miscounts when the tree is a DAG instead of a strict tree, by recursively building up the list of types of nodes as a hash and returning the node ids instead of the counts. That way they don't get double counted.
Also, adds history to time (teal) nodes and calculates a linear regression on that history to make a prediction about when the number of incomplete smalls goes to 0.