node: Implement checkpoints for recovery procedures

[goshawk-3]

Summary

In scenarios similar to the one outlined in https://github.com/dusk-network/rusk/issues/1093, it is essential to have the ability to revert to either the "last epoch" or the most recent "valid" finalized state.

Possible solution design or implementation

Currently, we have a function called try_revert, that supports reversion to a single target - LastFinalized. However, this function can be expanded to enable reversion to additional historical checkpoints. Together with that, we have to either automatically detect&resolve any unrecoverable state or allow the reversion via CLI.

[herr-seppia]

See also #970

[autholykos]

This should solve the persistence error popping up when improperly shutting down the node:

Error: PersistenceError(Custom { kind: InvalidData, error: "Invalid index file ""/opt/dusk/rusk/state/246607761c6c6a5b9791c08cbb38e1692eacee015b074864358e5c230f66d480/index"": context error: pointer out of bounds: base 0x7ffcbc0cad08 offset -68 not in range 0x5603bfdda080..0x5603bfdda0

[autholykos]

Depends on #1516

[HDauven]

Superseded by #1684