Cosmetic changes:
* Add 4 lines of comments (now every statement has a comment)
* Avoid unnecessary `assert` (the end goal for me is for `assert`s to be exclusive to the operation processing helpers).
* Merge `return`s into one (increase readability, reduce verbosity)
* Use shorter-named `bit_0_indices` and `bit_1_indices` helper variables
Substantive change:
* Remove the condition that `len(0_indices) + len(1_indices) > 0`. This condition is redundant in the context of `process_attester_slashing` because of `slashed_any`. It is largely artificial in `process_attestation` where validators are incentivised to maximise new attestations.
The state transition spec should be reasonably self-contained, limiting the amount of "magic" outside of it. This PR is a first step in this direction, specifically for operation processing.
Add 32-byte of arbitrary "graffiti" data in beacon blocks, in a similar vein to `extraData` in Eth1. To be used in wonderful and unpredictable ways (permissionless innovation by block proposers).
* Remove `get_crosslink_committees_at_slot` (that function's ugly man...)
* Make the "base" that everything works off instead be `get_crosslink_committee`
* Attestations store epoch, start shard and shard, no longer slot (slot can be calculated from the other three)
* Retaining start shard in attestations allows `get_attesting_indices` to peek much further back into the past, making it useful for slashings (Phase 1)
* Some two-layer-deep nested loops become one-layer-deep loops
* Initialisation bug: initial `effective_balance` be not greater than `MAX_EFFECTIVE_BALANCE`
* Hysteresis bug: do not prevent `effective_balance` to go from `MAX_EFFECTIVE_BALANCE - 1` to `MAX_EFFECTIVE_BALANCE`
