cache justified roots in state/rename attestation fields

[djrtwo]

addresses #742

Two things:

correct value for genesis justified roots

Right now, we just have ZERO_HASH for the genesis justified roots. This does not actually reflect the value of the root of the block at genesis. The value of the block at the genesis slot is looks like

block = get_empty_block()
genesis_state = get_genesis_state()
block.state_root = hash_tree_root(genesis_state)

Due to the normal chicken-and-egg state/block-root problem, we can't embed the genesis block root in as the initial justified roots when constructing the genesis state. If we want to enforce this, we can do the following in cache_state. This is nice because the value reflects an actual block root but it's not really necessary imo.

if state.justified_root == ZERO_HASH:
    state.justified_root = hash_tree_root(state.latest_block_header)
    state.previous_justified_root = state.justified_root

finalized root

Should we add another field finalized_root for symmetry? It is not needed in the state transition but might be nice to be available in state regardless of the number of slots since finality. (if greater than HISTORICAL_ROOTS_LENGTH you would no longer have it available in state)

[JustinDrake]

Should we add another field finalized_root

Good idea! At the very least we want an EVM2.0 opcode to fetch the finalized root.