In WT-12791 we now can clear incremental backup bits based on checkpoints and the extent lists. That improvement is very dependent on backup granularity size and it is hard to measure savings across runs. With that change, the blocks necessary for the existing checkpoints are the ones considered.

This ticket should investigate keeping information about the "old" way in memory at least and providing statistics to be able to compare how much data we are now copying versus "if we didn't clear any bits".

This new statistic could allow for experimentation with different granularity sizes.