docs(evaluate): Benchmarks page — selecting and running benchmarks

[sephmard]

Part of #1436 — Evaluate section restructure.

The current evaluation section has a Browse Environments page (environment-list.mdx) that lists all 90+ environments but does not orient readers to how to select a benchmark for evaluation, what makes a benchmark trustworthy, or how to interpret benchmark results.

Tasks

• Create or significantly revise to produce fern/versions/latest/pages/evaluation/benchmarks.mdx
• Cover:
  • What a benchmark is vs a training environment (evaluation-only vs dual-use)
  • How to select a benchmark for a target capability
  • Brief overview of benchmark categories (math, coding, reasoning, agentic, safety)
  • Link to the Browse Environments page for the full list
  • Link to Add a Benchmark (Contribute) for contributors
• Assess whether environment-list.mdx should remain as a sibling page or be merged/linked
• Add navigation card in evaluation index
• fern check passes