Backport "Inner hashing of value in state trie (runtime versioning)"

[tomaka]

paritytech/substrate#9732

[tomaka]

After #1970 and #1971, what remains is the implementation of the actual changes in the trie. And that's going to be complicated, especially without any specification.

[tomaka]

Some note for myself:

In the Substrate client, ext_storage_root_version_2 always passes as parameter the same value as in RuntimeSpec::state_version. It is RuntimeSpec::state_version that is authoritative as to which version to use for the storage writes. In smoldot I'm going to panic if ext_storage_root_version_2 gets passed a different value.

[tomaka]

The most pressing change to make at the moment is implementing the V1 trie format so that we can generate the proper genesis hash root even if state_version is equal to 1, which is important for the light client to be able to connect to some chains.
Implementing ext_storage_root_version_2 and all is more than secondary.

[tomaka]

https://spec.polkadot.network/#defn-node-header