Test fixture complexity creates cognitive load (unique HouseholdIds, feature flags, isolation)

[DigiBanks99]

Budget integration tests use:

• Unique HouseholdId prefixes per test (90..., 91..., etc.)
• Feature flags (Features:Budget, Features:BudgetItems)
• Two factory types (BudgetTestWebApplicationFactory vs TestWebApplicationFactory)
• JSON deserialization with PropertyNameCaseInsensitive
• CreateIsolatedFactory pattern for isolation

When writing a new test, need to understand all these patterns. Not documented together.

Recommendation: Create TESTING.md or add to existing docs with example of a 'typical test' showing all the patterns in context.