fix ppov2_trainer tensorboard logging bug

[DZ9]

Tensorboard log fails because current code doesn't update global step.

• before:
  

• after this PR:
  

• test command

python examples/scripts/ppo/ppo.py \
    --learning_rate 3e-6 \
    --output_dir ./output \
    --per_device_train_batch_size 1 \
    --gradient_accumulation_steps 1 \
    --total_episodes 100 \
    --model_name_or_path Qwen2/Qwen2-0.5B-Instruct \
    --non_eos_penalty \
    --sft_model_path Qwen2/Qwen2-0.5B-Instruct \
    --reward_model_path Qwen2/Qwen2-0.5B-Instruct \
    --report_to tensorboard \
    --local_rollout_forward_batch_size 1 \
    --logging_strategy steps \
    --logging_steps 1 

[HuggingFaceDocBuilderDev]

The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update.