test(critic-fixes): address brittleness, over-mocking, and missing edge cases (#20)

test(wave-4): LLM orchestration tests for eligibility, intake, and research (#19)