Re-measure, elaborate and publicize use of slower disks for colder state/data

We already [document](https://docs.erigon.tech/fundamentals/optimizing-storage) this, but we need to re-measure and expand. Rationale: requiring only a smallish but fast NVMe storage for hot data and offloading the rest to slower disks makes hardware requirements for running an Ethereum node more acceptable. Publicizing such results might nudge other clients into accepting elements of our data model.