fix: a non-dict finding silently drops all raw research findings (#1739)

tests/test_research_handler_raw_nondict.py

@@ -0,0 +1,14 @@
+from src.research_handler import ResearchHandler
+
+
+def test_extract_raw_findings_skips_non_dict_without_losing_all():
+    # The body is wrapped in a try/except that returns [] on any error, so a
+    # single non-dict finding made the AttributeError from f.get swallow EVERY
+    # good finding (silent total data loss), not just the bad row.
+    findings = [
+        {"url": "https://a.com", "summary": "a real and useful finding here"},
+        "junk-row",
+        {"url": "https://b.com", "summary": "another genuine finding with detail"},
+    ]
+    out = ResearchHandler._extract_raw_findings(findings)
+    assert [i["url"] for i in out] == ["https://a.com", "https://b.com"]