Add high reasoning evals for Claude, GPT-5

[addmorebass]

I am curious to see how much difference adding thinking makes for models like Sonnet 4.5 and GPT-5 makes on intelligent document processing

[github-actions]

👋 Welcome! Thanks for opening your first issue. If you'd like to take a crack at fixing it, feel free to open a pull request — otherwise, we'll take a look as soon as we can!