Optimize SimpleTextBKDWriter.rotateToTree tree-layout computation to O(1) per node

[rajat315315]

Fixes: #16353

Description

This PR optimizes the rotateToTree method in SimpleTextBKDWriter to avoid iterative loop overhead when calculating binary tree partitions.

rotateToTree converts a sorted, flat list of leaf block split/start values into a flat array structure representing a balanced complete binary search tree.

The Problem / Bottleneck

To partition nodes for the left subtree (leftHalf) at each level of recursion, the method previously used an iterative while(true) loop to calculate the level-by-level boundaries. This loop ran in O(log(count)) time at each step of the recursion. Since recursion occurs for every node in the tree, the overall mathematical partition computation overhead was O(N log N) for N inner nodes.

The Optimization

1. O(1) Left Subtree Count: We replaced the iterative while(true) loop with a bitwise constant-time O(1) calculation to determine the exact number of nodes in the left subtree of a complete binary tree:

   int h = Integer.highestOneBit(count);
   int halfH = h >>> 1;
   int leftHalf = halfH - 1 + Math.min(count - h + 1, halfH);

This leverages the fact that complete binary trees have a mathematically rigid structure that can be derived in constant time using bit-shifts.

3. Unified Array Copying: We consolidated the System.arraycopy code to handle both leaf and internal nodes under a single, unified execution path, reducing code duplication and branch prediction overhead.

4. Dead Code Cleanup: Removed legacy debug output and outdated comments.