One round poll to broker is needed to read the message.Second round poll to broker is needed to detect EOF.The test is asserting that we get message and EOF within 2s.But the consumer is configured with 'fetch.wait.max.ms': 1000.We don't fit 2 timout out polls within 2s.We need to increase the consume timeout to 3s.

One round poll to broker is needed to read the message.Second round poll to broker is needed to detect EOF.The test is asserting that we get EOF, data and EOFwhen data is produced at 2nd second.But the consumer is configured with 'fetch.wait.max.ms': 1000.We read the data in poll to broker slightly after 3rd second.We detect the 2nd EOF slightly after 4th second.We need to increase the consume timeout to 5s.

Steps to reproduce:* configure consumer with setDefaultConsumeTimeout(1000)* produce 1 message every 500ms* call consumer.consume(128, cb)Actual outcome:* consume returns 128 messages after 64 secondsExpected outcome:* consumer returns ~2 messages after 1 secondKafkaConsumerConsumeNum call underlaying c++ m_consumer->Consumein cycle until* either the accumulated batch is full* or the call to c++ m_consumer->Consume times out on the total timeoutKafkaConsumerConsumeNum must enforce its timeoutover all m_consumer->Consume invocations altogether.