[Feature Request] support dynamic sampling for GRPO trainer

### Feature request

Filter most easy and hard samples can increased the diversity of examples, which is beneficial for fine-tuning performance.

trl can provide a custom function between generation and computing per_token_logps for user to evaluate the quality of prompt, to determinate whether current sample be used in next step or not.

### Motivation

- dynamic sampling from [DAPO](https://arxiv.org/abs/2503.14476), page 5, 3.2 section.
- Rollout Rescue Mechanism and Intra-Batch Informative Substitution in [Polaris](https://hkunlp.github.io/blog/2025/Polaris/)

### Your contribution

I'm pleased to contribute this feature