Add types.Header marshaling using rlpgen

[piersy]

Previously rlpgen was used to generate faster rlp encoding methods for types.Header (Approx 40% faster). When updating to support data migrated from Celo we dropped the use of the generated code, but we should consider adding it back for the two header structures we support.

[palango]

@Kourin1996 this is fixed with #288, right?

[Kourin1996]

@palango I think so. This issue should be same with https://github.com/celo-org/celo-blockchain-planning/issues/720.