Ralph Loop 2026-03-16: Latest State Matched Research Handoff

Why

PRD-003 candidate review already exposed top_matched_research_candidate inside the live review packet, but the raw research/latest_state.json handoff still only persisted deployable_trade, review_candidate, and research_candidate at the top level. Any direct consumer of the handoff file could therefore lose the best semantically matched research row even when the candidate review builder had it.

Change

1. Updated research/continuous_loop.py so loop writes persist top-level matched_research_candidate alongside the other top candidate summaries.
2. Updated research/status.py hydration persistence so a status refresh also backfills matched_research_candidate into legacy or partial research/latest_state.json payloads.
3. Added focused regression coverage in tests/test_continuous_loop.py and tests/test_research_status.py.
4. Documented the raw handoff guarantee in research/README.md.

Validation

1. pytest -q tests/test_continuous_loop.py tests/test_research_status.py - 18 passed in 0.18s
2. python3 -m py_compile research/continuous_loop.py research/status.py - passed

Outcome

Raw latest-state consumers now get the same matched-research top candidate that status-oriented consumers already saw, which improves candidate-review handoff quality without changing any deployment gate or sleeve logic.