Some fixes to finality tests

[djrtwo]

Fix finality_12. This should be a good base to fix the others.

• need to set the associated epoch boundary block roots to ensure that get_matching_target_ works
• justification bit 0 should be set at first so it gets shift to position 1. Then 0 gets justified and you have 0 and 1 for finality.
• aggregation_bitfield -> aggregation_bits in add_mock_attestations (use bool array for init)
• add_mock_attestations had bug, setting target=source. fixed
• add_mock_attestations sets shard in attestation.data.crosslink. This piece of data is required to figure out what committee is attesting. Previous, all were defaulting to zero so each committee added looked the same.

[CarlBeek]

It feels strange to do this as it is not the behaviour one would expect from this test. The problem, however, is that otherwise, a new checkpoint is finalised as the condition of "2nd/3rd most recent epochs are justified, the 2nd using the 3rd as source" is met. This means that in the case where there is insufficient support, assert state.finalized_checkpoint == old_finalized # no new finalized fails. @djrtwo, please can you check me here (and propose an alternative solution if you have one.)

TLDR: this is required because of an artifact of how the tests are constructed by manipulating what is and isn't justified in the state.

[CarlBeek]

I extended these finalisation fixes to all of the finalisation cases. Fixes include:

• all of the boundry_epochs are now set (needed for the correct calculation of attestations).
• Many list-slicing bugs
• correction of what is justified in each of the tests.

[protolambda]

Fixed the attestation creation, since it was broken on mainnet config (multi committees per slot). And also reworded the bitfield indices and justification, it's clear now.