Parallel-Reasoning (Learning Adaptive Parallel Reasoning with Language Models)