docs: expand repository topics for structured-output discovery

[reacher-z]

Current state

The repository currently exposes only three GitHub topics: benchmark, evaluation, and structure. Those labels omit the terms researchers and coding/search agents normally use for this benchmark, especially structured outputs, LLM evaluation, and visual/code generation.

Proposed topics

Keep the existing topics and add this focused set:

• llm-evaluation
• llm-benchmark
• structured-output
• structured-generation
• large-language-models
• code-generation
• multimodal-evaluation
• visual-evaluation
• instruction-following
• benchmark-dataset
• json
• xml
• yaml

All describe documented StructEval capabilities or supported formats, and the combined total remains below GitHub’s 20-topic limit.

Acceptance criteria

• Add the proposed topics without removing the existing three.
• Verify the repository appears under the key topic pages after GitHub re-indexes it.
• Revisit the list if the benchmark’s supported formats or evaluation scope changes.

This issue was prepared with assistance from OpenAI Codex after a read-only audit of the public repository metadata. The attempted direct settings update was rejected for insufficient repository-settings permission, so no topics were changed.