[2.8.3 Backport] ISGR Pull checkpoint sequences unreliable

ISGR pull checkpoint sequences can be both behind or ahead of the actual intended sequence to be checkpointed.

This has the effect that pull replications that are interrupted (aborted or disconnected replications, both one-shot and continuous) can cause:

• Skip sequences entirely
• Reprocessed sequences

One-shot or caught up continuous replications may get a sequence behind the intended checkpoint sequence, but never ahead - resulting in only:

• Reprocessed sequences

Technical Detail

The checkpointer expects the slice of {{expectedSeqs }} for a replication to be ordered. In the pull case, this list of sequences comes from a map, which has random ordering, so does not produce an ordered set of sequences for the checkpoint calculation logic to work from, resulting in the checkpointer choosing the last sequence in the map iteration, which is not nessisarily the latest sequence.