Refactor account management in ntx-builder

[SantiagoPittella]

We measured the ntx-builder with a large account (1M storage map entries) and realized that the current approach with the full Account in-memory is not scalable.

We should try an alternative, treating the native accounts the same way the foreign-account path already does which is using the store as an oracle, and the ntx-builder holds only the header state in memory. It requires some changes to the database, but overall reduces the amount of code. In memory we would have a PartialAccount which isn't a problem long term. This approach also prevents the divergence that we had in the past of the in memory account not matching the node one, since the store is only updated by proven blocks and all the state is read from there.

[Mirko-von-Leipzig]

What is in PartialAccount? Just what gets loaded as one goes?

We could probably get very far by using only sqlite + caching layer in front of it. This would include caching the native faucet for example (which would happen organically).

[SantiagoPittella]

As defined in protocol:

pub struct PartialAccount {
    /// The ID for the partial account
    id: AccountId,
    /// Partial representation of the account's vault, containing the vault root and necessary
    /// proof information for asset verification
    partial_vault: PartialVault,
    /// Partial representation of the account's storage, containing the storage commitment and
    /// proofs for specific storage slots that need to be accessed
    partial_storage: PartialStorage,
    /// Account code
    code: AccountCode,
    /// The current transaction nonce of the account
    nonce: Felt,
    /// The seed of the account ID, if any.
    seed: Option<Word>,
}

And is used in the execution. In the current implementation we create this in every transaction from the full Account.