node_decommissioning_test: fix order of throttle and decommission

If we throttle after we issue the decom command, decom can finish before
the throttling will take effect. The correct order (if we want the
partition movements to get stuck) is throttle-then-decom.

tests/rptest/tests/nodes_decommissioning_test.py

@@ -584,11 +584,13 @@ def test_decommissioning_finishes_after_manual_cancellation(
         to_decommission = random.choice(self.redpanda.nodes)
         node_id = self.redpanda.node_id(to_decommission)
 
+        # throttle recovery
+        self._set_recovery_rate(100)
+
         survivor_node = self._not_decommissioned_node(node_id)
         self.logger.info(f"decommissioning node: {node_id}", )
         self._decommission(node_id)
 
-        self._set_recovery_rate(100)
         # wait for some partitions to start moving
         wait_until(lambda: self._partitions_moving(node=survivor_node),
                    timeout_sec=15,