Cleanup containers

[JustinDrake]

• Remove unnecessary per-field comments, focus on comments for field grouping
• Group Validator fields relevant to light-clients together (small optimisation for light clients)
• Rework grouping of BeaconState fields to "tell a story":
  • # Versioning (in space and time) specifies a blockchain
  • # History (block/state roots) is the "bones" of any blockchain
  • # Eth1 voting leads to deposits
  • # Registry emerges from deposits
  • # Shuffling the registry leads to committees
  • # Attestations are committee votes
  • # Crosslinks emerge from attestations
  • # Finality also emerges from attestations
• deposit_index => eth1_deposit_index

Do not merge before:

• Use custom types in data structures and constants #695 which specifies custom types for individual container fields, offsetting the removal of comments in some instances
• GENESIS_SLOT == 0 #896 and Make use of object initial values #843 so that we don't have to continuously maintain the genesis BeaconState
• add previous and current crosslinks #874 which introduces current_crosslinks and previous_crosslinks
• Possible aesthetic rework to get_domain #865 which renames proof_of_possession to signature

[djrtwo]

Not sure I agree with the grouping here. pubkey and slashed are relevant to far more than light clients

[dankrad]

So I kind of disagree with this. I believe it is good to have a human-readable interpretation of a field next to it, I just think the comments should be clearer than they are now. This goes exactly against what I propose in #942.

My suggestion is that in an ideal world, a field has a corresponding description that can be checked against the actual implementing algorithm. Let's give an example:

{
    # Validator indices
    'custody_bit_0_indices': ['uint64'],
    'custody_bit_1_indices': ['uint64'],
    # Attestation data
    'data': AttestationData,
    # Aggregate signature
    'aggregate_signature': 'bytes96',
}

I agree that the comment "Validator indices" is useless. However, if we change it to

{
    # Indices of all the validators that voted attested with custody bit = 0
    'custody_bit_0_indices': ['uint64'],
    # Indices of all the validators that voted attested with custody bit = 1
    'custody_bit_1_indices': ['uint64'],
    # Attestation data
    'data': AttestationData,
    # Aggregate signature
    'aggregate_signature': 'bytes96',
}

then it actually becomes more readable and we can use this information to check whether the algorithm computing this actually does what it claims.

[protolambda]

@JustinDrake
Now that #695 and #1077 are complete, can we get this PR ready again, if you still like to merge it? It needs an update to account for the new fields, and the syntax is effectively just name: type now.

[JustinDrake]

Close in favour of #1156