Add ImageBench V1 to Benchmark Suites

[dh7]

Adds ImageBench V1 to Section 9 → Benchmark Suites.

Fit note. The existing suite entries (LiveBench, lm-evaluation-harness, HELM, SWE-bench, VLMEvalKit, etc.) are all LLM / multimodal-understanding benchmarks. ImageBench adds the complementary angle: a prompt-adherence benchmark for text-to-image generation models, judged by a VLM panel.

Why it fits "opensource AI":

• Full prompt + per-model + per-judgment dataset is public on Hugging Face: https://huggingface.co/datasets/dh7/imagebench (CC0 — reproducible from the raw data)
• Methodology page is fully public
• All model outputs are rendered side-by-side on public head-to-head compare pages
• 60+ frontier models covered (Flux 2 Max, Nano Banana Pro, Seedream V4, Reve 2.1, Imagen 4 Ultra, Qwen Image 2.0 Pro, and more)

Happy to reword, slot elsewhere (e.g. under Evaluation Frameworks or a new Image Generation Benchmarks subsection), or close if it's not a fit for this list.