RecursionCircuit and updated BatchCircuit

[noel2004]

This PR induce a new aggregation layer to aggregate multiple snarks from batch circuit, using an recursive scheme:

doc

[lispc]

1. is the bundle config file here is used for production? if not, there should be some code to let recursion circuit use "layer5" config, where is that code?

2. withdraw root is "Additional public input that are unchanged over all the SNARKs". not too clear to me. You mean these are public inputs not to be "chained"?

[noel2004]

1. is the bundle config file here is used for production? if not, there should be some code to let recursion circuit use "layer5" config, where is that code?
2. withdraw root is "Additional public input that are unchanged over all the SNARKs". not too clear to me. You mean these are public inputs not to be "chained"?

1. If indicate the bundle_circuit.config file under aggregator/config: No. This file is only used by the unit test of recursion circuit, so we need 5 adv colums for this config since one unit test would recursive prove a recursion circuit itself. And maybe we should not naming it as "bundle" i think.

Recursiion circuit (bundle) use a "layer5" config here:

zkevm-circuits/prover/src/aggregator/prover.rs

Lines 229 to 236 in 8846bd4

	let layer5_snark = self.prover_impl.load_or_gen_recursion_snark(
	&name,
	LayerId::Layer5.id(),
	LayerId::Layer5.degree(),
	&bundle_snarks,
	output_dir,
	)?;

2. I think it is not "unchanged" but "do not involved in state transition". I consider these PIs are "not require to be verified by state transition", because they are in fact involved with (or say, decided by) state root. The doc has some irriadiation about them:
   https://www.notion.so/scrollzkp/Upgrade-4-Darwin-Documentation-05a3ecb59e9d4f288254701f8c888173?pvs=4#62a0b2058b9842188b5bd5e7267220ca

[lispc]

ok i find this line env::set_var("BUNDLE_CONFIG", layer_config_path(id));

[roynalnaruto]

Great work @noel2004 🎉