test: integrate fork choice test vectors

[unnawut]

What was wrong? & How was it fixed?

I just spotted that we have the fork choice test vectors implemented but they're not being run. After enabling them there were 3 main groups of failures:

1. Fork choice tie-breaking should be based on voting weight then the lexical order of the hash. But the code uses (vote_weight, slot, *child_hash) so it's ordered by weight, then slot number first before the hash. In test cases where it intentionally puts 2 blocks of equal votes, but the block at lower slot has a higher lexical block, the test failed. --> This is fixed by removing slot from the tuple.
2. When a new block is processed by on_block(), the fork choice store collects the attestation data from block.body.attestations into attestation_data_by_root_provider, but it doesn't collect from BlockWithAttestation::proposer_attestation so in test vectors where it's quite normal to have a block with only the proposer's attestation and no non-proposer attestation, the fork choice store won't have the attestation data, and so won't aggregate the proposer's attestation and so the weight is never updated. --> Fixed by adding proposer's attestation data to attestation_data_by_root_provider
3. The test vector setup uses Signature::blank() while the test vectors are using real signatures now. -> Fixed by downloading the keys from https://github.com/leanEthereum/leansig-test-keys and use them in lean-spec-tests

Note that there are 2 changes to the fork choice store logic, but I think at most it would make the weights stronger so hopefully this doesn't break our devnet participation.

To-Do

• I have read CONTRIBUTING.md.
• This PR title follows conventional commits.

[unnawut]

triggers an actual tick so attestations get aggregated