Backport is_parent_strong and is_head_weak changes to Phase0

[mkalinin]

Backports is_parent_strong and is_head_weak Gloas code to Phase0, specifically:

• Replace get_weight with get_attestation_score which in the case of get_proposer_head shouldn’t change the result because the difference between the two is whether proposer boost is counted or not which in the case of proposer reorg isn’t relevant because the head is late and no proposer boost is applied
• Count head committee equivocating validators towards head support

There seems to be no reason for these changes to be not applied since Phase0. cc @potuz

[jtraglia]

Count head committee equivocating validators towards head support

If implemented now, is there any risk in this behavior change causes issues on mainnet?

[mkalinin]

Count head committee equivocating validators towards head support

If implemented now, is there any risk in this behavior change causes issues on mainnet?

i don't think so. it might result in clients handling proposer reorg a bit differently which should be fine as it doesn't affect attesting at all; moreover, differences in its handling could only happen in an edge case where a decent amount of head's committee stake is slashed

[jtraglia]

LGTM. I would appreciate reviews from clients (eg @nflaig, @potuz, @michaelsproul) before merging though.

[michaelsproul]

As far as I recall the equivocating balance stuff is a bit hacky in our Gloas code (and the source of compliance test failures). We might rework our impl and apply the fixes pre-Gloas, or we might just keep it confined to post-Gloas for simplicity.

We are planning to delete pre-Gloas fork choice as soon as the fork is live, so I'm reluctant to put more engineering work into something that will be deleted so soon.