LedgerDB namespacing

[jfldde]

Description

• Split up LedgerDB operations by usage
• Introduces SharedLedgerOps for any method shared between Prover/Sequencer/Node.
• Adds NodeLedgerOps, ProverLedgerOps and SequencerLedgerOps to constraint the associated CitreaFullNode, CitreaProver and CitreaSequencer

Linked Issues

• Fixes LedgerDb namespacing for sequencer, node, prover #911

[rakanalh]

I just gonna leave a comment here without "requesting changes" since i don't agree with the way things have been done here but i don't want to block the PR either.

The reason i have is:
We have Node,Prover,SequencerTrait... etc which are all implemented for LedgerDB which doesn't make that much of a difference from the previous version which LedgerDB had by containing all the methods. Limiting the access to methods using traits in prover for example by defining ledger_db: DB is okay.

The way i would have done it is as follows:
One trait, 3 structs

/// Common ledger ops, leaving out the word `common`/`shared` from the name.
pub trait Ledger {
  ... common stuff
}

pub struct ProverLedger {
}
impl ProverLedger {
}
impl LedgerOps for ProverLedger {
}

Same for sequencer and fullnode. I would also try to namespace tables

sequencer_xyz
prover_abc`

so that prover tables are only accessed within ProverLedger.

[jfldde]

The main aim of this PR is to restrict visibility in order to prevent calling the method where it shouldn't be, i.e. calling a sequencer method from the prover.
I think we would have the same end result with the three structs you're describing.

If I understand correctly, we would declare the Ledger trait with default implementation that would then be implemented by the three structs (they have the exact same implementation for now).

For the namespacing of the tables, that would be ideal, but would it change anything if we're sharing the inner RocksDB instance ?

[jfldde]

Isn't the three structs approach basically trying to do OOP where the Ledger trait would be the base class ?
Is that the paradigm we tend to follow ?
Curious on the approach we can take for namespacing the tables.