Add model benchmark scores (MMLU, HumanEval, etc.)

[i-need-token]

Context

The AI Models Catalog tracks pricing, context windows, and capabilities. One frequently requested feature is benchmark scores — MMLU, HumanEval, MATH, GPQA, etc.

What to do

1. Add a benchmarks optional field to the model schema in types/model.ts
2. Add the Zod validation in types/schemas.ts
3. Add a few benchmark entries for well-known models as examples
4. Update the interactive catalog to show benchmark scores
5. Create a docs/benchmarks.md page

Suggested schema

benchmarks:
  mmlu: 88.7
  humaneval: 92.0
  math: 78.3
  gpqa: 65.2

Notes

• Benchmark data should come from official model cards/reports (first-party)
• Not all models have benchmarks — this is an optional field
• Follow the pattern of existing optional fields