kafka/s: Reject leader epoch requests in future

- If the user attempts to make a offset_for_leader_epoch request where
the requested epoch is ahead of the latest known, the response should
contain -1s for requested epoch and end_offset.

- Fixes: redpanda-data#11864

src/v/kafka/server/replicated_partition.cc

@@ -339,6 +339,11 @@ replicated_partition::get_leader_epoch_last_offset_unbounded(
  const model::term_id term(epoch);
  const auto first_local_offset = _partition->raft_start_offset();
  const auto first_local_term = _partition->get_term(first_local_offset);
+ const auto last_local_term = _partition->term();
+ if (term > last_local_term) {
+ // Request for term that is in the future
+ co_return std::nullopt;
+ }
  // Look for the highest offset in the requested term, or the first offset
  // in the next term. This mirrors behavior in Kafka, see
  // https://github.com/apache/kafka/blob/97105a8e5812135515f5a0fa4d5ff554d80df2fe/storage/src/main/java/org/apache/kafka/storage/internals/epoch/LeaderEpochFileCache.java#L255-L281