feat: Epoch orchestrator

[spalladino]

Paused in favor of #8675

[PhilWindle]

I think it will require the VK. Maybe it doesn't now because we don't recursively verify the AVM proof, but we will do.

[PhilWindle]

I'm not sure I follow. All of the public kernels are recursively verified at the moment.

[spalladino]

Yeah, but public kernels require not just the recursive proof and vk, but also the vkPath, vkIndex, and clientIvc:

aztec-packages/yarn-project/circuits.js/src/structs/kernel/public_kernel_data.ts

Lines 21 to 41 in a098d41

	/**
	* Proof of the previous kernel.
	*/
	public proof: RecursiveProof<typeof NESTED_RECURSIVE_PROOF_LENGTH>,
	/**
	* Verification key of the previous kernel.
	*/
	public vk: VerificationKeyData,
	/**
	* Index of the previous kernel's vk in a tree of vks.
	*/
	public vkIndex: UInt32,
	/**
	* Sibling path of the previous kernel's vk in a tree of vks.
	*/
	public vkPath: Tuple<Fr, typeof VK_TREE_HEIGHT>,
	/**
	* TODO(https://github.com/AztecProtocol/aztec-packages/issues/7369) this should be tube-proved for the first iteration and replace proof above
	*/
	public clientIvcProof: ClientIvcProof = ClientIvcProof.empty(),

However, in the current orchestrator, only proof and vk are set:

aztec-packages/yarn-project/prover-client/src/orchestrator/tx-proving-state.ts

Lines 87 to 89 in a098d41

	// pass both the proof and verification key forward to the next circuit
	nextFunction.publicKernelRequest.inputs.previousKernel.proof = proof;
	nextFunction.publicKernelRequest.inputs.previousKernel.vk = verificationKey;

[PhilWindle]

I'm having difficulty understanding the general flow. I can see here we update the nested proof for each base parity, what kicks off the root parity once the final base parity is completed?