[wu-sheng]

Fix the consumer race in spring-kafka-1.3.x and 2.2.x scenarios

Follow-up to #841, which fixed this race in spring-kafka-2.3.x-scenario. The 1.3.x and 2.2.x scenarios have the same code, so they have the same latent race. They still pass in CI only because they usually win it.

Cause

• The consumer has no committed offset and starts from the latest offset.
• healthCheck returned 200 as soon as Tomcat started.
• If the entry request sent helloWorld before the consumer got its partition, the consumer skipped the message. The request then waited forever on the latch, so its segment was never reported (SegmentNotFoundException for GET:/case/spring-kafka-case).

Fix

• healthCheck reports healthy only after a ConsumerRebalanceListener sees a non-empty assignment, the same as the 2.3.x fix.
• This leaves no window. The scenarios disable auto commit, so the container calls consumer.position() and commits for every assigned partition before it calls the user listener. I checked this in the bytecode of 1.3.9, 1.3.10, 1.3.11 and 2.2.14.

Verification

• I ran every version in both support-version.list files locally against a fresh Kafka 2.4.1 broker, with the mock collector, the agent and the validator.
• The health check is polled every 0.2s, and the entry is called right after it passes.
• group.initial.rebalance.delay.ms stands in for a slow runner.

Version	main's controller, 15s delay	This PR, 3s delay	This PR, 15s delay
1.3.9.RELEASE	FAILED (entry segment missing)	PASSED	PASSED
1.3.10.RELEASE	FAILED (entry segment missing)	PASSED	PASSED
1.3.11.RELEASE	FAILED (entry segment missing)	PASSED	PASSED
2.2.14.RELEASE	FAILED (entry segment missing)	PASSED	PASSED

At the broker's default 3s delay, main's controller failed once in two 1.3.11 runs. The health check now returns 500 until the assignment, about 3s or 15s.

This changes only test scenarios; there is no plugin or CHANGES.md change.