feat(base_layer/core): add domain hashing wrapper for consensus encoding

[sdbondi]

Description

• adds DomainSeparatedConsensusHasher
• renames ConsensusHashWriter to ConsensusHasher
• removes Write implementation from ConsensusHasher
• adds TransactionHashDomain hash domain
• uses domain hashing for transaction metadata signature and script challenges

Motivation and Context

Creates an ergonomic and 0-alloc hasher for consensus encoded values.

Usage:

  DomainSeparatedConsensusHasher::<TransactionHashDomain>::new("metadata_signature")
            .chain(public_commitment_nonce)
            .chain(script)
            .chain(features)
            .chain(sender_offset_public_key)
            .chain(commitment)
            .chain(covenant)
            .chain(encrypted_value)
            .finalize()

How Has This Been Tested?

New unit test, existing tests use hashing

[hansieodendaal]

This looks nice.

I think one of the issues with the failing integration tests is the hash calculation in

• buildMetaChallenge(
• buildScriptChallenge(
• hashOutput(features, commitment, script, sender_offset_public_key)

from what I could pick up. I was busy chasing some other failures after ^^ have been corrected before I closed #4373.

[hansieodendaal]

Sorry, finger trouble closing this

[sdbondi]

@hansieodendaal lol no problem :) Oh right, I'll need to update cucumber tests. Thanks

[sdbondi]

Cucumber metadata signature is still broken, it seems to be the encrypted value is incorrect in the transactionBuilder. Going to leave that for another PR.

[CjS77]

LGTM. See comment.

[CjS77]

Couldn't we provide a blanket Write impl on the underlying DomainHasher?

[sdbondi]

In this case we can't because the API for Write is meant to allow "chunked" writes (network sockets, files etc) and the DomainSeparatedHasher does not allow this.

c = a||b
H.chain(a).chain(b) != H.chain(c)

This is the main reason why I chose to make the write interface internal to the hasher, because we know that ConseususEncoding writes each element in its entirety and our internal Write implementation never tells the caller to wait to write.