fix(core)!: consensus hashing without extraneous length prefixes for each write

[sdbondi]

Description

• Removes extraneous length prefixes for each write done in consensus encoding.
• cucumber: Corrects hashing for metadata, kernel and script sigs in cucumber tests
• cucumber: Adds missing minimumValuePromise value from metadata hash
• cucumber: fix use of hex string instead of buffer in hash
• new genesis for esme and igor
• new faucet utxos for esme

Motivation and Context

Because the H.chain(a).chain(b) != H(a||b) property does not hold in domain-separated hashes, we cannot allow writes to call Digest::update on a domain hasher.

Prefixing the length for every element is redundant with ConsensusEncoding as it is canonical and well-defined (either fixed length, length prefixed, bit-prefixed (varint)).

The PR uses a hash domain (impl DomainSeparation) to prepend a domain label, but uses consensus encoding directly after that so that H.chain(a).chain(b) != H(a||b) holds.

Implementing the hashing in other languages, as in the cucumber tests, now becomes fairly trivial.

How Has This Been Tested?

Clear out mempool cucumber used for testing and passes confirming that all javascript hashes match the rust backend.