Magma should create checkpoints only on data size

Currently we force create checkpoints every two minutes or 2% of data size. In slower workloads, this results in very high wasted space as observed in CBSE-16847. To clear away wasted space Magma will have to perform additional compaction increasing Write Amplification.

We can prevent this by removing the 2 minute checkpointing for rollback and only hold and create checkpoints based on overhead and data size.

We will have to evaluate the impact of this on tombstone purging, forced compaction, DCP scans and rollback.