ForkedHashedBeaconState helpers

[arnetheduck]

Using func avoids a template expansion at every usage site and just looks nicer .. unlike earlier nim versions, good code is generated for lent.

An additional benefit is that mutability semantics are preserved and the resulting value is immutable even when being accessed indirectly (unlike addr+[] which results in a var).

No matter what, when iterating over validators we continue to be affected by:

• lent not respected with for loop and array values nim-lang/Nim#25470
• Conversion from distinct in for forces a copy of underlying instance nim-lang/Nim#25469

[tersec]

One nice aspect of the slight ugliness of getStateField is that it is somewhat involved:

nimbus-eth2/beacon_chain/spec/forks.nim

Lines 987 to 1020 in f896feb

	template withState*(x: ForkedHashedBeaconState, body: untyped): untyped =
	case x.kind
	of ConsensusFork.Gloas:
	const consensusFork {.inject, used.} = ConsensusFork.Gloas
	template forkyState: untyped {.inject, used.} = x.gloasData
	body
	of ConsensusFork.Fulu:
	const consensusFork {.inject, used.} = ConsensusFork.Fulu
	template forkyState: untyped {.inject, used.} = x.fuluData
	body
	of ConsensusFork.Electra:
	const consensusFork {.inject, used.} = ConsensusFork.Electra
	template forkyState: untyped {.inject, used.} = x.electraData
	body
	of ConsensusFork.Deneb:
	const consensusFork {.inject, used.} = ConsensusFork.Deneb
	template forkyState: untyped {.inject, used.} = x.denebData
	body
	of ConsensusFork.Capella:
	const consensusFork {.inject, used.} = ConsensusFork.Capella
	template forkyState: untyped {.inject, used.} = x.capellaData
	body
	of ConsensusFork.Bellatrix:
	const consensusFork {.inject, used.} = ConsensusFork.Bellatrix
	template forkyState: untyped {.inject, used.} = x.bellatrixData
	body
	of ConsensusFork.Altair:
	const consensusFork {.inject, used.} = ConsensusFork.Altair
	template forkyState: untyped {.inject, used.} = x.altairData
	body
	of ConsensusFork.Phase0:
	const consensusFork {.inject, used.} = ConsensusFork.Phase0
	template forkyState: untyped {.inject, used.} = x.phase0Data
	body

i.e. being a kind of explicit call helps discourage a bunch of getStateFields near each other on the same state, while it would be better to have one withState covering them all. This PR hides the ugliness but it's a factory for spawning lots of unnecessary separate case statements/expressions on the fork.

[arnetheduck]

factory for spawning lots

yeah I thought about this but I think it's negligible - specially given the fact that all fields are at the same offset meaning that effectively, the case should go away, or almost at least. If there was any significant computation involved, I'd certainly agree.

[arnetheduck]

also, we don't use Forked* in tight loops anyway, typically .. ie everywhere where performance matters, we go Forked

[github-actions]

Unit Test Results

       12 files  ±0    2 440 suites  ±0   51m 2s ⏱️ +23s
12 669 tests ±0    9 192 ✔️ ±0    3 477 💤 ±0  0 ❌ ±0 
63 860 runs  ±0  51 430 ✔️ ±0  12 430 💤 ±0  0 ❌ ±0 

Results for commit edece56. ± Comparison against base commit f896feb.