Apply proposer boost if dependent roots match

[mkalinin]

Follow up to #5300.

Switches from proposer_index to dependent_root check when applying proposer boost. There is an edge case where proposer_index is the same on the two competing chains with different dependent roots, in this case applying PB to the diverged chain can be exploited. This PR is a fix for that edge case.

There are two ways how dependent_root can be computed: 1) with the block tree, root and store time 2) with the post-state of a block. (1) since we do this computation in the FC context.

Might worth making get_dependent_root an outer function.
UPD: made get_dependent_root a top-level function.

[michaelsproul]

Looks correct 👌

[jihoonsong]

Looks good!

[jtraglia]

Nice, I like this too. We'll need more client approvals before this can be merged though.

[jtraglia]

@mkalinin we should remove the following function & update the gossip condition to use the one introduced here.

consensus-specs/specs/gloas/p2p-interface.md

Lines 467 to 475 in 55519bb

	```python
	def get_proposer_dependent_root(state: BeaconState, epoch: Epoch) -> Root:
	"""
	Return the dependent root for the proposer lookahead at ``epoch``.
	"""
	return get_block_root_at_slot(
	state, Slot(compute_start_slot_at_epoch(Epoch(epoch - MIN_SEED_LOOKAHEAD)) - 1)
	)
	```

[jihoonsong]

I think it's better to handle the deprecation of get_proposer_dependent_root() in the Gloas executable p2p PR, which has more context around its usage. See my comment. @jtraglia

[jtraglia]

I think it's better to handle the deprecation of get_proposer_dependent_root() in the Gloas executable p2p PR

Okay yeah, that works too!

[mkalinin]

I think it's better to handle the deprecation of get_proposer_dependent_root() in the Gloas executable p2p PR, which has more context around its usage. See my comment. @jtraglia

Note that get_dependent_root will be returning Root() for the first two epochs. It might be important for devnets that are starting from Gloas. So, builders will need to use that function to construct bids which is reasonable, but need to be kept in mind

[jtraglia]

@mkalinin happy to merge this whenever you feel it's appropriate. I think we should probably get more reviews from clients. But I have a feeling this will be universally accepted as a good change.

[jihoonsong]

I think it's better to handle the deprecation of get_proposer_dependent_root() in the Gloas executable p2p PR, which has more context around its usage. See my #5294 (comment). @jtraglia

Note that get_dependent_root will be returning Root() for the first two epochs. It might be important for devnets that are starting from Gloas. So, builders will need to use that function to construct bids which is reasonable, but need to be kept in mind

Good point. Note that get_proposer_dependent_root() will underflow for the first two epochs.

[jtraglia]

Going to merge this. This was originally @potuz's idea here, @michaelsproul expressed support here, and @nflaig has expressed support via private chats. This is good enough for me. I don't expect this to be controversial.