merkle tree: investigate divergance between go/rust implementation

[liamsi]

It seems a completely empty tree is handled differently in Rust (all zeroes output) than in go (returns nil):

func HashFromByteSlices(items [][]byte) []byte {
	switch len(items) {
	case 0:
		return nil

https://github.com/tendermint/tendermint/blob/b5f030892d1f33dfffb7a3d108f63583316befe2/crypto/merkle/tree.go#L9-L21
vs.

tendermint-rs/tendermint/src/merkle.rs

Line 24 in 667fc8d

0 => [0; HASH_SIZE],

---

Also notice that this edge-case is treated differently in google's rfc6962 implementation (result of hashing an empty byte slice as the root hash of an empty tree):
https://github.com/google/trillian/blob/7166f485ebc965d24116db667c9e058a8ed9275c/merkle/rfc6962/rfc6962.go#L51

(This is true for both, the go and the rust implementation)

It's probably a good idea to follow the rfc6962 spec (other than we are compatible to that spec):

The hash of an empty list is the hash of an empty string:

MTH({}) = SHA-256().

https://tools.ietf.org/html/rfc6962#section-2.1

ref: tendermint/tendermint#4241 (comment)
ref: #73
ref: tendermint/tendermint#5192