MINOR: Fix testRackAwareAssignment flake

The last part of testRackAwareAssignment was found to be flaky. This
part moves all topic partitions to different racks and waits for
consumer assignments to settle. Each of the three consumers is expected
to revoke all its partitions and be assigned partitions previously held
by another within a 15 second timeout.

This timeout is not always sufficient. The consumer heartbeat interval
is left at the default of 5,000 ms and each consumer polls every
3,000 ms. In the worst case, it takes a consumer around 7,000 ms to
reconcile an assignment change. An additional 3,000 ms round of polling
may be required when a consumer needs to auto-commit offsets. Two rounds
of reconciliation must happen within 15,000 ms.

The timeline of an example failing run looks like:

-02.956 Group coordinator computes target assignment at epoch=6
        consumer0=[0] consumer1=[1, 2] consumer2=[3, 4, 5]
+00.000 15 second timeout starts

+03.179 consumer0 heartbeats
        This is the first heartbeat since the rack reassignments.
+03.179 Group coordinator computes target assignment at epoch=7
        consumer0=[5] consumer1=[3, 4] consumer2=[0, 1, 2]
+03.186 consumer0 heartbeat receives assignment []

+04.151 consumer1 starts poll()
+04.877 consumer1 heartbeats
+04.878 consumer1 heartbeat receives assignment []
+05.155 consumer1 ends poll()

+07.259 consumer1 starts poll()
+07.259     consumer1 sends auto-commit with offsets for [1, 2]
+07.288     consumer1 receives auto-commit response
+08.263 consumer1 ends poll()

+10.379 consumer1 starts poll()
+10.379     consumer1 calls onPartitionsRevoked with [1, 2]
+10.379     consumer1 calls onPartitionsAssigned with []
+10.382     consumer1 heartbeats with owned partitions []
+10.387     consumer1 heartbeat receives assignment [3, 4]
+10.483     consumer1 calls onPartitionsAssigned [3, 4]
+10.483     consumer1 heartbeats with owned partitions [3, 4]
+11.384 consumer1 ends poll()

+15.000 15 second timeout elapses and the test fails
+15.300 consumer2 heartbeat receives assignment [0, 1, 2]

Author: squah-confluent

clients/clients-integration-tests/src/test/java/org/apache/kafka/clients/consumer/ConsumerIntegrationTest.java

@@ -352,7 +352,7 @@ public void testRackAwareAssignment(ClusterInstance clusterInstance) throws Exec
                 return consumer0.assignment().equals(Set.of(new TopicPartition(topic, 5))) &&
                     consumer1.assignment().equals(Set.of(new TopicPartition(topic, 3), new TopicPartition(topic, 4))) &&
                     consumer2.assignment().equals(Set.of(new TopicPartition(topic, 0), new TopicPartition(topic, 1), new TopicPartition(topic, 2)));
-            }, "Consumer with topic partition mapping should be 0 -> 5 | 1 -> 3, 4 | 2 -> 0, 1, 2");
+            }, 30000, "Consumer with topic partition mapping should be 0 -> 5 | 1 -> 3, 4 | 2 -> 0, 1, 2");
         }
     }