core: Wait for backoff timer on address update in pick_first

The backoff timer is only used when serializeRetries=true, and that
exists to match the old/current pick_first's behavior as closely as
possible. InternalSubchannel.updateAddresses() would take no action when
in TRANSIENT_FAILURE; it would update the addresses and just wait for
the backoff timer to expire.

Note that this only impacts serializeRetries=true; in the other cases we
do want to start trying to the new addresses immediately, because the
backoff timers are in the subchannels.

Note that this change was also important because requestConnection() can
be directly triggered by the user with channel.getState(true), and that
shouldn't defeat the backoff timer.

Author: ejona86

core/src/main/java/io/grpc/internal/PickFirstLeafLoadBalancer.java

@@ -496,7 +496,7 @@ private void shutdownRemaining(SubchannelData activeSubchannelData) {
    */
   @Override
   public void requestConnection() {
-    if (!addressIndex.isValid() || rawConnectivityState == SHUTDOWN) {
+    if (!addressIndex.isValid() || rawConnectivityState == SHUTDOWN || reconnectTask != null) {
       return;
     }
 

core/src/test/java/io/grpc/internal/PickFirstLeafLoadBalancerTest.java

@@ -1446,6 +1446,11 @@ public void updateAddresses_disjoint_transient_failure() {
     loadBalancer.acceptResolvedAddresses(
         ResolvedAddresses.newBuilder().setAddresses(newServers).setAttributes(affinity).build());
 
+    if (serializeRetries) {
+      inOrder.verify(mockSubchannel3, never()).start(stateListenerCaptor.capture());
+      forwardTimeByBackoffDelay();
+    }
+
     // subchannel 3 still attempts a connection even though we stay in transient failure
     assertEquals(TRANSIENT_FAILURE, loadBalancer.getConcludedConnectivityState());
     inOrder.verify(mockSubchannel3).start(stateListenerCaptor.capture());