Revised AI engineering docs (2)

[dominicchapman]

New "evaluation" content:

• Updated workflow language from "Measure" to "Evaluate" to better reflect our approach
• Reorganized evaluation content into a dedicated section with six focused pages (overview, setup, write evaluations, flags & experiments, run evaluations, analyze results)

Other changes:

• Concepts: Added definitions for flags and experiments; integrated AI capability architecture spectrum (single-turn → workflows → single-agent → multi-agent)
• Create: De-emphasized experimental prompt management features while clarifying Axiom's current focus on evaluation and observability; added references to Vercel AI SDK examples and Mastra as framework alternatives
• Iterate: Complete rewrite introducing the systematic improvement loop; added sections on user feedback capture and domain expert annotation workflows (marked as coming soon); reorganized failure categorization by severity for better prioritization
• Quickstart: Updated to reference evaluation framework and CLI authentication; improved "What's next" guidance