In the previous timeseries write path, we could have some stats updated more eagerly if there is a bucket without all batches committed but we have called rollover, because we updated the stats in determineRolloverReason.

After SERVER-102319, we update the hard-closed rollover stats only when we know that we have rolled-over a bucket with all committed all of our measurements in the bucket (if we are attempting to rollover a bucket that doesn't have all measurements committed, we just set the rolloverReason of the bucket without any further action) within rollover.

made more general