Add AIME 2026 dataset support

[nkalyanv99]

Summary

• Add AIME 2026 dataset loading (MathArena/aime_2026) and preprocessing to data_utils.py
• 30 problems, integer answers — same schema as AIME 2024

Type of Change

• New feature (non-breaking change which adds functionality)

Testing

• All existing tests pass when running uv run pytest locally
• New tests have been added to cover the changes

Checklist

• My code follows the style guidelines of this project as outlined in AGENTS.md
• I have performed a self-review of my own code
• My changes generate no new warnings

---

Note

Low Risk
Small, additive change limited to example dataset loading/preprocessing; main risk is minor breakage if the upstream dataset schema or split name differs.

Overview
Adds AIME 2026 support to the example dataset utilities by wiring aime2026 into load_example_dataset (loads MathArena/aime_2026, defaulting to the train split).

Introduces a matching get_preprocess_fn handler that maps problem → question and normalizes answers to integer strings (str(int(...))) for consistent question/answer output.

^{Written by Cursor Bugbot for commit 3b55563. This will update automatically on new commits. Configure here.}

[mikasenghaas]

lgtm