Fix: list_snapshots mit -r statt -d1, Cache-Key pro Dataset

-d 1 begrenzte auf direkte Snapshots des Datasets, tank/share wurde
nicht eingeschlossen. -r (recursive) liefert alle Sub-Datasets.
Cache-Key jetzt dataset-spezifisch um Kollisionen zu vermeiden.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

backend/services/zfs_runner.py

@@ -362,16 +362,16 @@ class ZFSRunner:
         """
         List snapshots (with limit for performance on many snapshots)
         """
-        cached = self.cache.get("snapshots")
+        cache_key = f"snapshots:{dataset_name or '*'}"
+        cached = self.cache.get(cache_key)
         if cached:
             return cached
 
-        # If no dataset specified, list all
         if dataset_name:
-            cmd = ["zfs", "list", "-t", "snapshot", "-d", "1", "-H", "-p",
+            # No -d limit so sub-datasets (e.g. tank/share) are included
+            cmd = ["zfs", "list", "-t", "snapshot", "-r", "-H", "-p",
                    "-o", "name,used,referenced,creation", dataset_name]
         else:
-            # Get all snapshots, limited
             cmd = ["zfs", "list", "-t", "snapshot", "-H", "-p",
                    "-o", "name,used,referenced,creation"]
 
@@ -402,7 +402,7 @@ class ZFSRunner:
         snapshots.sort(key=lambda x: x["creation"], reverse=True)
         snapshots = snapshots[:limit]
 
-        self.cache.set("snapshots", snapshots, ttl_seconds=60)
+        self.cache.set(cache_key, snapshots, ttl_seconds=60)
         return snapshots
 
     def create_snapshot(self, dataset_name: str, snapshot_name: Optional[str] = None) -> Dict[str, str]: