Configure chain quality grace period for berkeley release

[nholland94]

This PR configures berkeley's grace period to 4.5 days, and ensures that the grace period works in hard forks.

The original code was hard coding the grace period in terms of a number of days, which would make the grace period unsafe under some network configurations. I exposed the grace period constant as a configuration in terms of slots, and ensured that the grace period end is accounting for enough slots in order to have a fresh sub window samples of the chain density. I also assert that the specified grace period is safe.

I also fixed a bug in the grace period logic where we weren't accounting for hard forks correctly. We were comparing the global slot since hard fork against a global slot since genesis when testing for the grace period. I fixed this and improved the types to add clarity.

I've confirmed the behavior manually by running a local network with 1/3rd of the stake online. The grace period end was computed as expected and the density only dropped after it was expected to.

[nholland94]

!ci-build-me

[nholland94]

!ci-build-me

[nholland94]

!ci-build-me

[nholland94]

!ci-build-me

[nholland94]

!ci-build-me

[nholland94]

!ci-build-me

[sebastiencs]

@nholland94 That new grace_period_slots is ignored in Protocol_constants_checked.to_input, is that expected ?

If 2 Protocol_state.Body have a different value of grace_period_slots, they will still get the same hash (when calling Protocol_state.Body.hash)

mina/src/lib/mina_base/protocol_constants_checked.ml

Lines 69 to 76 in d6cc26f

	let to_input (t : value) =
	Array.reduce_exn ~f:Random_oracle.Input.Chunked.append
	[| T.to_input t.k
	; T.to_input t.delta
	; T.to_input t.slots_per_epoch
	; T.to_input t.slots_per_sub_window
	; Block_time.to_input t.genesis_state_timestamp
	|]

[nholland94]

@sebastiencs That's a mistake, thanks for catching it! The to_input function should be destructuring it's input in order to avoid these kind of bugs. I'll put in a fix soon.