Separate header type and storage type

[svyatonik]

GRANDPA and parachains finality pallets are accepting the chain header/head and storing it directly in the runtime. For most of cases (including ours) we may only store the storage root, thus decreasing the occupied runtime storage + message transactions bandwidth (?).

The implementation may look like:

trait HeaderArtifacts {
  type Artifacts;
  fn into(header: Header) -> Self::Artifacts;
}

struct StorageRootArtifact;
impl HeaderArtifacts for StorageRootArtifact {
  type Artifacts = H256;
  fn into(header: Header) -> H256 {
    *header.storage_root()  
  }
}

...

impl pallet_grandpa::Config for Runtime {
 ...
 type Artifacts = StorageRootArtifact;
 ...
}

[svyatonik]

I'm prioritizing this issue, since we're now changing the "scope" of our pallets to specific case. So would be better to have this asap.

The same thing may be done for the parachains pallet (would require head parser configuration for every parachain)