In AccumulatedDataLoop miscounting - do not assume batches are all BlockBatchMaxSizeBytes size

[danwt]

Even after

• patch for mtsitrin 954 multiple accumulateddata in case of submission latency #962
• fix(blockManager): multiple accumulateddata trigger #960

which fix

• sequencer is failing to close the gap between the batch submitted height and the block produced height #959
• multiple AccumulatedData in case of submission latency #954

there is still a problem:

The AccumulatedDataLoop basically assumes that the size of the batch being submitted is equal to BlockBatchMaxSizeBytes but actually it will be less. That means every time the submission happens it is actually falling more and more behind the actual produced blocks.