[BENCHMARK] Add LongTextBench benchmark

[davidberenstein1957]

Description

Add LongTextBench for evaluating text-to-image with long prompts.

Details

• Dataset: X-Omni/LongText-Bench
• Collate: prompt_with_auxiliaries_collate

Implementation

• Add setup_long_text_bench_dataset in src/pruna/data/datasets/prompt.py
• Register in base_datasets
• Add BenchmarkInfo entry with metrics: ["text_score"]
• Add test

Acceptance Criteria

• PrunaDataModule.from_string("LongTextBench") works
• Auxiliaries include text field (for evaluation)
• Test passes