Undesired partition migration happening because of stale metadata

[emasab]

Description

A partition migration can happen, using stale metadata, when the partition is validating a next_fetch_start and retring the validation.
It can use stale metadata with an invalid leader epoch and migrate back to the previous leader. Later this is corrected but it's sub-optimal.

How to reproduce

Execute test 0146/do_test_stale_metadata_doesnt_migrate_partition in #4680.

Checklist

Please provide the following information:

• librdkafka version (2.1.0)
• Apache Kafka version: <REPLACE with e.g., 0.10.2.3>
• librdkafka client configuration: <REPLACE with e.g., message.timeout.ms=123, auto.reset.offset=earliest, ..>
• Operating system: <REPLACE with e.g., Centos 5 (x64)>
• Provide logs (with debug=.. as necessary) from librdkafka
• Provide broker log excerpts
• Critical issue