Remove v0 parachains runtime

[rphmeier]

All of the functionality except for the Registrar module has been superseded by the v1 runtime modules. I removed the Registrar regardless, leaving the runtime without parachains

The Registrar Module

The registrar module previously had these functions:

1. Parachain and Parathread registrar
2. Parachain and parathread scheduling
3. Parathread auctions

Function 1 is now mostly provided by the Paras module. This Registrar should provide a small wrapper around the Paras for handling deposits.
Function 2 is now handled by the Scheduler module.
Function 3 should be handled by a separate a parathread_auction module.

So although the Registrar is currently being removed, those relevant pieces of logic can be recovered from the commit history when the new version is created in a follow-on PR.

We need to add para swapping (one feature of the old registrar) to the Paras module and make some small tweaks: swapping should only happen at session boundaries.

Compatibility with previous runtime

There are two main compatibility concerns: Call indices and SessionKey layout. For the former, I replaced each of the existing parachains modules with an instanced dummy module. For the latter, I will ensure that the session key layout remains the same by keeping the Parachains session key around, but where updates are simply discarded for the time being.

I also created a dummy implementation of the v0::ParachainHost runtime API. This can be removed once the BlockBuilder in polkadot-validation no longer relies on it.

[rphmeier]

@bkchr I'm getting a runtime error:

conflicting implementations of trait `sp_api_hidden_includes_construct_runtime::hidden_include::IsSubType<polkadot_runtime_common::dummy::Call<Runtime>>` for type `Call`:

probably because there is no typed differentiator of the dummy instances.

I'm also getting "duplicate lang item in crate std" which is puzzling, since I didn't add or change any dependency of the polkadot-runtime crate or the polkadot-runtime-common crate in this PR.

[rphmeier]

Icing until we can test the runtime upgrades.

[gui1117]

Registrar, Parachains, and Slots have some keys on chain (in polkadot), I guess they should be removed with some migration code
(maybe kusama also have some attestation keys)

[rphmeier]

I guess they should be removed with some migration code

I think we can also defer to governance on this, right? But we could also do remove_prefix

[gui1117]

I did start a kusama and a polkadot testnet with storage datas from kusama and polkadot (except babe, grandpa, session and staking). with the runtime from this PR (except with small session length and era length) and the client from v0.8.19, and it worked fine finalizing and new eras.

still I think we need to adress the origin changing issue otherwise looks good to me

[gui1117]

This changes the outer origin enum, this requires an upgrade of scheduler::Agenda storage (or a dummy variant in origin, maybe using Dummy)

(scheduler agenda holds the origin since paritytech/substrate#6387 )

[rphmeier]

Great point. We'll add the parachains origin to this PR

[rphmeier]

We did some digging, and found that the migration is not really necessary because:

1. The council on Polkadot and Kusama has not scheduled anything with a Parachains origin and will not, either
2. The Origin enum had the Parachains origin at the end, so no indices of other origins were altered