Remove parameterized-consensus-hook feature

[s0me0ne-unkn0wn]

Closes #4366

[skunert]

I am trying to reason about the implications of these change.

After this PR is merged, all parachain runtimes using polkadot-sdk of next release need to supply a ConsensusHook. Since we have async backing enabled everywhere soon, this makes sense, but still a breaking change.

ConsensusHook Overview

• We have ExpectParentIncluded. This is a workaround to make collators work that don't have parent_head included in their relay chain state proof. This was introduced in august 2023, I think we should deprecate this consensus hook in this PR. Longterm remove this logic here.
• We have RequireParentIncluded and FixedCapacityUnincludedSegment. These don't perform any additional validation on the relay chain storage proof and set a fixed capacity to the unincluded segment. This one is still useful if you don't want the slot checking logic.
• We have FixedVelocityConsensusHook which should be used for aura-based chains. It does slot validation logic and has a fixed unincluded segment length.

Which brings me to the deprecation of CheckInherents. Parachains were able to supply their own implementation of the trait. I am wondering if there are use cases where a parachain has custom inherents that they were able to validate using CheckInherents, but not using the ConsensusHook.

A quick github search shows that most chains are doing only slot checks that are now living in FixedVelocityConsensusHook. Usage of this seems in general very inconsistent, for example its not used in any of the system-parachains. @bkchr I remember we discussed briefly about the CheckInherents and its deprecation, do you remember the outcome?

[bkchr]

@bkchr I remember we discussed briefly about the CheckInherents and its deprecation, do you remember the outcome?

The hook gets the StateProof passed. So, essentially the same as CheckInherents. Even if someone would have a custom implementation (which I doubt), they can still do the same checks as before. (Okay, a little bit harder if you need the transaction, but generally not impossible).

[skunert]

(Okay, a little bit harder if you need the transaction, but generally not impossible).

I mean that was the interesting part, I know that the state proof goes into the hook 👍 . But I agree that there are probably no custom implementations.