Fix out-of-bounds in get_shuffling

[paulhauner]

What

Change get_shuffling(...) so it gives shuffles using an index of active_validator_indices instead of a value of active_validator_indices.

Why

The following fails with a IndexError: list index out of range:

validator_indices = [
    0,    # inactive
    1,    # active
    2,    # inactive
    3,    # active
]

# ... therefore ...

active_validator_indices = [
    1,    # active
    3,    # active
]

def get_permuted_index(i, length):
    return i

shuffled_active_validator_indices = [
        active_validator_indices[get_permuted_index(i, len(active_validator_indices))]
        for i in active_validator_indices
]

Whereas this does not:

shuffled_active_validator_indices = [
        active_validator_indices[get_permuted_index(i, len(active_validator_indices))]
        for i in range(len(active_validator_indices))
]

[JustinDrake]

@djrtwo Are we no longer seeding with epoch?

[djrtwo]

We moved epoch into the mix in generate_seed rather than on the fly in shuffle. This was clearer and still accomplishes the goal of ensuring that the shuffling changes at points of reshuffling even in the case that no new randao reveals occur.

[djrtwo]

nice catch!