Switch to Haiku model for cost-effective AI scaling

[ytsarev]

Description of your changes

Summary

Migrated intelligent RDS scaling from Sonnet to Haiku model, achieving 85%+ cost reduction on AI inference while maintaining decision quality.

Changes

• Updated CronOperation to use claude-haiku-4-5 model (operations/rds-intelligent-scaling-cron)
• Added automated e2e tests for deterministic part of the Configuration
• Updated test credentials and configurations for v2 namespaced resources
• Simplified prompt - thanks to reliable markdown sanitization, no verbose output constraints required anymore

Measured Cost Impact

Measured cost reduction: 85-89%

• Sonnet baseline: ~$45-90/day
• Haiku: ~$9.52/day

At 1-minute cron schedule (demo frequency):

• Monthly savings: ~$1,500-2,400 per database
• Annual savings: ~$18,000-29,000 per database

For production schedules (5-10 minute intervals), costs scale proportionally down while maintaining intelligent scaling benefits.

Dependencies

Requires function-claude v0.4.0 for reliable Haiku markdown sanitization.

I have:

• Read and followed Upbound's contribution process.
• Run make reviewable to ensure this PR is ready for review.
• Added backport release-x.y labels to auto-backport this PR, as appropriate.

How has this code been tested

• Composition tests

 up test run tests/*
  ✓   Parsing tests
  ✓   Collecting resources
  ✓   Generating language schemas
  ✓   Checking dependencies
  ✓   Building functions
  ✓   Building configuration package
  ✓   Pushing embedded functions to local daemon
  ✓   Assert test-mysql
  ✓   Assert test-postgres
SUCCESS:
SUCCESS: Tests Summary:
SUCCESS: ------------------
SUCCESS: Total Tests Executed: 2
SUCCESS: Passed tests:         2
SUCCESS: Failed tests:         0

• E2E Tests

 up test run tests/* --e2e

Full Validation

✅ Acceptance tests passing with Haiku model
✅ AI scaling decisions remain accurate and well-reasoned
✅ Clean event messages (markdown handling)
✅ Production metrics confirm 85%+ cost reduction