Remove temporary band-aid introduced via #1781. Once we upload state checkpoints to the object store, they are accessible to all partition processor replicas. Hence, it is theoretically possible to trim the log once the checkpoint was stored. However, if some replicas are lagging behind, this can lead to them encountering trim gaps. So either, we wait until all replicas have reached the lsn of the checkpoint before trimming or we add support to handle trim gaps by downloading the latest checkpoint.