[feat] tests should be format agnostic ie: no regex test at runtime

[mclenhard]

Refactored 4 regex-heavy evals to use runtime testing:

• 031-ai-sdk-migration-simple: Now renders components and tests actual behavior (11 regex →
  runtime tests)
• 001-server-component: Calls async server component and verifies fetch/render (10 regex →
  4 runtime tests)
• 022-prefer-server-actions: Renders form and checks DOM attributes (14 regex → 5 runtime
  tests)
• 030-app-router-migration-hard: Imports modules and tests exports/rendering (25 regex → 7
  runtime tests)