feat(investigation): preserve cumulative reference graphs

Accumulate investigation paths and preserve failed attempts across rounds. Keep graph annotations separate from execution state, synchronize the API contract, and consolidate documentation.

tests/support/operational_task_repository.py

@@ -104,6 +104,19 @@ class TaskRepository:
     async def get_latest_progress(self, task_id, generation):
         return self.progress.get((task_id, generation))
 
+    async def get_latest_graph_progress(self, task_id):
+        snapshots = [value for (stored_id, _), value in self.progress.items()
+                     if stored_id == task_id
+                     and (value["progress"].get("diagram") or {}).get("edges") is not None]
+        return max(snapshots, key=lambda row: (row["execution_generation"], row["revision"])) if snapshots else None
+
+    async def get_latest_diagram(self, task_id):
+        diagrams = [(row["published_at"], row["progress"].get("diagram"))
+                    for (stored_id, _), row in self.progress.items() if stored_id == task_id]
+        diagrams.extend((row["published_at"], row["report"].get("diagram")) for row in self.reports)
+        return next((diagram for _, diagram in sorted(diagrams, key=lambda item: item[0], reverse=True)
+                     if diagram is not None), None)
+
     async def save_progress(self, task_id, snapshot):
         self.progress[(task_id, snapshot["execution_generation"])] = snapshot