Does provable absence of barren plateaus imply classical simulability? Or, why we need to rethink variational quantum computing.pdf