When the ChunkManager is updated due to a split, we want to take whatever size estimates we have currently for a chunk and divide it among the split chunks so that we don't lose our current estimate or underestimate the number of bytes in the resulting chunks.