Yield in group state recovery loop

[mmaslankaprv]

When a consumer group is being recovered the whole log state is read and cached in a non preemptive loop this may cause reactor stalls. The loop should yield to scheduler instead