Rewrite Tree With Subtree Sums

You are given the roots of two complete binary trees, root1 and root2. Problem Overview You are given two complete binary trees named root1 and root2. These two trees are identical in shape. They have the exact same structure and the same number of nodes. Your task is to update the values inside root2. For every node in root2, you must calculate the sum of the matching subtree in root1. This sum should include the node in root1 itself and all of its descendants (children, grandchildren, etc.). After updating all the values, return the modified root2. Sample Cases Example 1 Input: root1 = [5,2,3,1,4,6,7], root2 = [9,9,9,9,9,9,9] Output: [28,7,16,1,4,6,7] Explanation: To find the new values for root2, we calculate the subtree sums from root1: Leaf Nodes: The nodes at the bottom (1, 4, 6, 7) have no children. Their subtree sum is just their own value. Node 2: This node has children 1 and 4. The sum is 2 + 1 + 4 = 7. Node 3: This node has children 6 and 7. The sum is 3 + 6 + 7 = 16. Root Node 5: This node includes the sums from the left side (7) and the right side (16). The total is 5 + 7 + 16 = 28. Example 2 Input: root1 = [1,2,3,4,5,6], root2 = [0,0,0,0,0,0] Output: [21,11,9,4,5,6] Input Limitations Tree Size: The number of nodes is between 1 and 100,000 ($10^5$). Node Values: The value of each node is between -10,000 and 10,000. Tree Type: Both root1 and root2 are complete binary trees. Structure: root1 and root2 have the exact same shape. Notes Snowflake frames this as "compute each node's subtree sum and write it onto the corresponding node of a second tree," which is a single post-order DFS: subtree_sum(node) = node.val + subtree_sum(left) + subtree_sum(right), assigning into the matching node of the output tree on the way back up. The follow-ups push on parallelism: if there are very many subtrees and only a limited number of worker threads, how do you speed up the aggregation? Sibling subtrees are independent, so a bottom-up post-order can be parallelized — dispatch ready (leaf-most) subtrees onto a fixed-size thread pool and join results at each parent, ideally with work-stealing so idle threads pick up any subtree whose children have finished. The thread count is bounded by hardware, not by the number of subtrees. Edge cases: single-node tree, skewed tree (recursion depth — switch to an explicit stack or the iterative post-order for very deep inputs).