refactor(triedb): add clean cache to the rawDB engine

[MaksymMalicki]

This PR should be reviewed after #3187

• This PR is the first of series of PRs, which introduce components of the pathDB into the rawDB. They are meant to be reviewed and benchmarked separately, to measure the impact on the new state performance.
• This PR introduces a clean cache layer for the rawDB. The state is meant to read from this layer before the direct PebbleDB hit. We write to clean cache on read miss and also when flushing the nodes to the DB.

Next PR to review: #3192