Evaluation Toolkit Guide

[psschwei]

Description:
Guide to using the evaluation and tuning toolkit for systematic agent improvement.

Acceptance Criteria:

• Guide at docs/guides/evaluation.md
• When and why to evaluate agents
• Writing effective test cases
• Running evaluations (programmatic and CLI)
• Interpreting results and metrics
• Parameter sweeps: when and how
• Prompt comparison workflow
• CI/CD integration patterns
• Cost management for evaluations

[psschwei]

This list was put together from the old version of the docs. Now that we've done a thorough revision, probably best to start anew on what we need to add.