bcachefs: Fix lock ordering with snapshot_create_lock

We must not hold btree locks while taking snapshot_create_lock - this
fixes a lockdep splat.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

fs/bcachefs/snapshot.c

@@ -1447,6 +1447,7 @@ int bch2_delete_dead_snapshots(struct bch_fs *c)
 		}
 	}
 
+	bch2_trans_unlock(trans);
 	down_write(&c->snapshot_create_lock);
 
 	for_each_btree_key(trans, iter, BTREE_ID_snapshots,