Fix flake in ManagedChannelImplTest.transportFailsOnStart

The scheduling on another thread led to a race where sometimes the
future wasn't completed by the time isDone() was checked in
ClientCallImpl causing the usage of DelayedStream, which really messed
up what the test was trying to do.

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

@@ -150,16 +150,7 @@ final class TransportSet {
     Preconditions.checkState(!shutdown, "Already shut down");
     Preconditions.checkState(reconnectTask == null || reconnectTask.isDone(),
         "previous reconnectTask is not done");
-    long delayMillis;
+    Runnable createTransportRunnable = new Runnable() {
-    if (reconnectPolicy == null) {
-      // First connect attempt
-      delayMillis = 0;
-      reconnectPolicy = backoffPolicyProvider.get();
-    } else {
-      // Reconnect attempts
-      delayMillis = reconnectPolicy.nextBackoffMillis();
-    }
-    reconnectTask = scheduledExecutor.schedule(new Runnable() {
       @Override
       public void run() {
         synchronized (lock) {
@@ -177,7 +168,18 @@ final class TransportSet {
               "failed to set the new transport to the future");
         }
       }
-    }, delayMillis, TimeUnit.MILLISECONDS);
+    };
+    if (reconnectPolicy == null) {
+      // First connect attempt
+      reconnectPolicy = backoffPolicyProvider.get();
+      createTransportRunnable.run();
+      reconnectTask = null;
+    } else {
+      // Reconnect attempts
+      long delayMillis = reconnectPolicy.nextBackoffMillis();
+      reconnectTask = scheduledExecutor.schedule(
+          createTransportRunnable, delayMillis, TimeUnit.MILLISECONDS);
+    }
   }
 
   /**
@@ -196,7 +198,9 @@ final class TransportSet {
       if (transports.isEmpty()) {
         runCallback = true;
       }
+      if (reconnectTask != null) {
         reconnectTask.cancel(false);
+      }
       // else: the callback will be run once all transports have been terminated
     }
     if (savedActiveTransportFuture != null) {