feat: support time-based forking

[leruaa]

Closes #937

[gakonst]

Let's discuss this before merging

[leruaa]

@gakonst yeah I made some impacting changes...

And yet I held back, I was this close to replace:

reth/crates/primitives/src/lib.rs

Line 65 in 78ffd0a

pub type BlockNumber = u64;

by

pub struct BlockNumber(u64);

[mattsse]

pub struct BlockNumber(u64);

what's the motivation for this? this would be super inconvenient imo.

edit: to distinguish between timestamp and blocknumber?

[mattsse]

moving this to ChainSpec, seems like the correct change to me.

@Rjected can review these changes best in detail

there are several unwraps and the Default inits that look a bit weird to atm.

[leruaa]

edit: to distinguish between timestamp and blocknumber?

Exactly, to be able to impl From<BlockNumber> for ForkDiscriminant and From<Timestamp> for ForkDiscriminant

[onbjerg]

Seems ok overall, not sure what I would have done much differently. We need to figure out what we are going to do about the places where we hard depend on Paris having a set block number; this is not the case for all networks, so we should prefer TTD.

[Rjected]

good start! most notes are about MergeNetsplit - this has unfortunate naming because it's basically only used in sepolia, and ttd is still used in mainnet

[leruaa]

@Rjected I think I've taken all change requests into account, the PR can be reviewed again

[Rjected]

just a few nits, otherwise lgtm!

[mattsse]

lgtm

pending @Rjected

[Rjected]

🚀 lgtm! thank you!

[gakonst]

I think this introduced a regression.

getting a lot of Failed to validate header error=HeaderValidation { hash: 0xb4fbadf8ea452b139718e2700dc1135cfc81145031c84b7ab27cd710394f7b38, error: TheMergeDifficultyIsNotZero }

This happens all the time. @onbjerg probably correctly identified that this is incorrect:

reth/crates/primitives/src/chain/spec.rs

Lines 147 to 148 in e8d7c05

	self.paris_ttd <= Some(discriminant.total_difficulty) ||
	*block_number <= Some(discriminant.block_number)

because of how it's called here:

reth/crates/consensus/src/verification.rs

Lines 53 to 55 in e8d7c05

	ForkDiscriminant::ttd(
	chain_spec.terminal_total_difficulty().unwrap_or_default(),
	Some(header.number),

Should we revert this PR? And investigate a proper solution, in preparation for Shanghai? cc @onbjerg @mattsse @Rjected

Also see this WIP spec ethereum/EIPs#6122