[BENCHMARK] Add OneIG benchmarks (Alignment + Text Rendering)

[davidberenstein1957]

Description

Add OneIG benchmarks including alignment benchmarks for anime, human, and object generation, plus text rendering evaluation.

Details

OneIG Alignment

• Dataset: OneIG-Bench/OneIG-Bench + GitHub JSON for questions
• Subcategories: anime_alignment, human_alignment, object_alignment
• Collate: prompt_with_auxiliaries_collate

OneIG Text Rendering

• Source: GitHub CSV from OneIG-Bench/OneIG-Benchmark repo
• Collate: prompt_with_auxiliaries_collate

Implementation

• Add setup_oneig_alignment_dataset in src/pruna/data/datasets/prompt.py
• Add setup_oneig_text_rendering_dataset in src/pruna/data/datasets/prompt.py
• Support category param for filtering (Anime_Stylization, Portrait, General_Object)
• Register in base_datasets
• Add BenchmarkInfo entries with appropriate metrics
• Auxiliaries should include questions, dependencies, category, text_content fields
• Add tests

Acceptance Criteria

OneIG Alignment

• PrunaDataModule.from_string("OneIGAlignment") works
• PrunaDataModule.from_string("OneIGAlignment", category="anime") works
• Auxiliaries include questions, dependencies, category fields
• Test passes

OneIG Text Rendering

• PrunaDataModule.from_string("OneIGTextRendering") works
• Auxiliaries include text_content field
• Test passes