perf(qualys/dashboard): parallelise fetch vulns par batch IPs (8 workers, ~18min -> ~3min)

app/services/qualys_service.py

@@ -959,18 +959,25 @@ def compute_vuln_dashboard(db, triggered_by="manual", run_id=None):
         """)).fetchall()
         asset_count = len(rows)
 
-        # 5. Recuperer vulns par batch de 50 IPs (cache 10min via get_vuln_counts)
+        # 5. Recuperer vulns par batch de 50 IPs en parallele (cache 10min via get_vuln_counts)
+        # Une session par thread pour eviter les conflits SQLAlchemy
+        from app.database import SessionLocal
         ip_to_vuln = {}
         unique_ips = list({str(r.ip_address) for r in rows
                            if r.ip_address and str(r.ip_address) != "None"})
-        for i in range(0, len(unique_ips), 50):
+        batches = [unique_ips[i:i+50] for i in range(0, len(unique_ips), 50)]
-            batch = unique_ips[i:i+50]
+        def _fetch_vuln_batch(batch):
+            s = SessionLocal()
             try:
-                vmap = get_vuln_counts(db, ",".join(batch))
+                return get_vuln_counts(s, ",".join(batch))
+            except Exception:
+                return {}
+            finally:
+                s.close()
+        with ThreadPoolExecutor(max_workers=8) as ex:
+            for vmap in ex.map(_fetch_vuln_batch, batches):
                 if vmap:
                     ip_to_vuln.update(vmap)
-            except Exception:
-                pass
 
         # 6. Classifier + agreger en utilisant les associations API
         agg = {}