feat: reconcile deleted yaml services

[vigneshrajsb]

Summary

• add YAML ownership metadata on deployables so stale reconciliation only touches YAML-owned services
• reconcile deleted/renamed YAML services behind features.reconcileDeletedServices
• add DeployCleanupService for deploy-scoped infra cleanup and stale service row deletion
• wire an async deploy cleanup queue for future infra-only cleanup callers
• skip deleted-service reconciliation for selective service redeploys by passing skipDeletedServiceReconciliation through the API redeploy, resolve queue, and build queue path

Safety

• service renames are treated as delete + create; no rename inference
• repo-filtered reconciliation is limited to matching resolvedFromRepositoryId
• ambiguous repositoryId = null deployables are skipped during repo-filtered reconciliation
• full config reconciliation may delete YAML-owned repositoryId = null deployables after YAML import succeeds
• DB-primary, merged DB/YAML, and configuration deployables are excluded from stale cleanup
• deploy cleanup uses deploy-scoped resource names/selectors only; no namespace/build-wide cleanup helpers
• selective service redeploys are not treated as authoritative deleted-service reconciliation scopes

Tests

• generalized BuildService stale reconciliation tests for:
  • feature flag disabled
  • deleted YAML-owned service cleanup
  • renamed YAML service cleanup
  • repo-filtered cleanup preserving other repository scopes
  • full cleanup of YAML-owned repositoryId = null services
  • repo-filtered skip for ambiguous null repository ownership
  • non-authoritative YAML import skip
  • YAML-owned non-configuration query scope
  • cleanup failure still deleting stale DB rows
  • selective service redeploy skipping stale reconciliation
  • selective service redeploy preserving the skip flag through resolve/build queues
• generalized DeployCleanupService tests for:
  • deploy-scoped Kubernetes, Helm, secret, and CLI cleanup
  • no service-name-only selector or build-wide cleanup usage
  • infra mode marks deploy torn_down only on full success
  • partial infra failure leaves status unchanged
  • service mode continues after partial teardown failure
  • stale service DB rows are deleted in one transaction

Validation

• pnpm run lint passed
• pnpm exec jest src/server/services/__tests__/build.test.ts --runInBand passed
• pnpm run test -- src/server/services/__tests__/build.test.ts src/server/services/__tests__/deployCleanup.test.ts src/server/services/__tests__/deployable.test.ts passed
• BUILD_MODE=yes DATABASE_URL=postgresql://lifecycle:lifecycle@localhost:5434/lifecycle APP_DB_HOST=localhost APP_DB_PORT=5434 APP_DB_USER=lifecycle APP_DB_PASSWORD=lifecycle APP_DB_NAME=lifecycle APP_DB_SSL=false pnpm run test passed
• pnpm run ts-check was run and still fails with repo-wide existing TypeScript strictness errors unrelated to the stale reconciliation changes
• pnpm exec jest --runInBand was run after the final redeploy-path fix and failed in this shell because many suites require DATABASE_URL / APP DB env; focused regression tests passed