Currently, state is committed to the block and state store using asynchronous calls followed by a synchronous call at the end.

We need to investigate whether explicitly batching them improves performance. In particular, consider the work done in :
tendermint/tendermint#6067
tendermint/tendermint#6018

and issue: tendermint/tendermint#5888

The goal of this is two fold:

• Increase atomicity guarantees
• Improve performance
  DoD:

• Understand, from a DB perspective whether atomicity is actually increased or not
• Evaluate whether there are corner cases where the batch becomes very big and understand how the DB handles this underneath (memory vs. disk)
• State is saved in batches with an indication on performance improvement