refact: use last path instead of key id to shorten the PathMessage

The current PathMessage contains 28 bytes as follows:

struct PathMessage {
    /// Reserved at this moment.
    reserved: [u8; 3],
    /// The random 'w' of Pedersen commitment `r*G + w*H`.
    w: u64,
    /// The key identifier. 1-byte path depth and 16-bytes BIP-32 key derivation path.
    key_id: Identifier,
}

The complete Identifier here is a little bit wasting on the precious chain space, this PR shorten the key_id here, replace it with the last_path which is only 4 bytes instead of 17 bytes.

pub struct PathMessage {
    /// The random 'w' of Pedersen commitment `r*G + w*H`
    pub w: i64,
    /// The last path index of the key identifier
    pub key_id_last_path: u32,
}

This is a balance we have to make, between a compact chain space and a simpler wallet implementation. The new PathMessage only need 12 bytes.

For wallet restoring or checking, to get a complete key id, the wallet must know the parent key path, which should be bound to the wallet account (i.e. m/p1/p2).

For example the default wallet account:

$ gotts-wallet --floonet account

____ Wallet Accounts ____

 Name    | Parent BIP-32 Derivation Path 
---------+-------------------------------
 default | m/0/0

After this PR, the wallet will never be able to restore all coins by one call, since a wallet could have multiple wallet accounts. It will have to leave to the wallet user to manage the wallet accounts manually, but I think it's acceptable, considering the original intention of the "accounts".

Additional note: At maximum, a Gotts key id (i.e Identifier) can have 4-level BIP-32 derivation path: m/p1/p2/p3/p4. The 1st two path consist of the so-called wallet account here. The last path can be saved into the chain into the PathMessage here. Then, how about the 3rd path?

At this moment, let's leave a TODO here about this 3rd path. In my current design (including some refactoring on this part soon), I just leave this 3rd path always as 0. This 3rd path could be used in the future if there's a good use case, but since we already have 2^31 keys here for each m/p1/p2/0, it's already fair enough for most of the use cases.

(Because this PR is a consensus breaking change, I also import a new Genesis block for Floonet test.)

gottstech / gotts

refact: use last path instead of key id to shorten the PathMessage #37