Adding rollout filtering to trainer

[LuanBrt]

Description

This PR adds rollout filtering to the Trainer, following the RAGEN approach. It helps prevent model collapse by retaining only the top N% of tasks with the highest outcome uncertainty.

Type of Change

• [ ] Bug fix (non-breaking change which fixes an issue)
• [x] New feature (non-breaking change which adds functionality)
• [ ] Breaking change (fix or feature that would cause existing functionality to not work as expected)
• [ ] Documentation update
• [ ] Test improvement

Testing

• [x] All existing tests pass
• [ ] New tests have been added to cover the changes
• [x] Tests have been run locally with python -m pytest tests/

Test Coverage

• Current coverage: 100%
• Coverage after changes: 100%

Checklist

• [x] My code follows the style guidelines of this project
• [x] I have performed a self-review of my own code
• [x] I have commented my code, particularly in hard-to-understand areas
• [x] I have made corresponding changes to the documentation
• [x] My changes generate no new warnings
• [x] Any dependent changes have been merged and published

Additional Notes

[CLAassistant]

All committers have signed the CLA.

[LuanBrt]

hey @willccbb, given the current uptades on verifiers, this feature can still be considered desirable to be included in the Trainer? Just to know if I should close the PR, otherwise I will try to make it compatible with the latest version.