Apply the finding engine to schematic and PCB reviews.

FACT, REQUIREMENT, and INFERENCE are separate fields; datasheet provenance
lives in the rule DB so Recommended cannot stay ERROR; LLM output is REVIEW.
Designer decisions persist across rescans. Schematic and PCB share one library
and the same finding object in the report UI.

tests/test_finding_schema.py

@@ -57,6 +57,29 @@ def test_new_fields_round_trip_json():
     assert again.variant == "DNP"
 
 
+def test_engine_fields_round_trip():
+    f = Finding(
+        designator="U1",
+        finding="x",
+        status="WARNING",
+        facts="net GND",
+        requirement="PSEL high",
+        inference="intentional?",
+        provenance="RECOMMENDED",
+        finding_class="RISK",
+        confidence=0.4,
+        evidence_status="SUFFICIENT",
+        calculation="n/a",
+        assumptions=["firmware unknown"],
+        action="leave PSEL low",
+    )
+    again = Finding.model_validate(json.loads(f.model_dump_json()))
+    assert again.facts == "net GND"
+    assert again.provenance == "RECOMMENDED"
+    assert again.finding_class == "RISK"
+    assert again.assumptions == ["firmware unknown"]
+
+
 def test_unknown_extra_keys_do_not_break_legacy_payloads():
     f = Finding.model_validate(
         {