logs22:Refactoring ideas

• Fix an issue where remove_generated is necessary.
• Running auth twice stuck.
• Remove train_with_reward
• All tests pass.
• Rename train_with_reward2 to train_with_reward
• Cleanup rl optimizer part.
• Refactoring train_rl_new and run it every Mode == test
• Save and restore in the train_rl
• Test save and restore
• See if pre-training and explore works.
• Identify the root cause of 6700 steps halt
  • Logging delta time per step