test: actors: manual Ni-PoRep onboarding

[rvagg]

Also adds multi-miner support to the manual test, including async PoST scheduling.

This depends on a bunch of changes elsewhere to work and needs manual wiring up:

• Ni-PoREP (temp, working branch for testing) builtin-actors#1546 which I've included here as an actors bundle (DO NOT MERGE WITH THIS)
• feat: miner: niporep params and cbor encoding vector tests go-state-types#270
• fix: s/interactivec/interactive for 512 NI-PoRep constants rust-filecoin-proofs-api#103
• Ni-PoRep Go/cgo additions filecoin-ffi#456
• Non Interactive Proof of Replication ref-fvm#1999 rebased on master to add some stuff and remove bls pairing; also some awkward manual wiring I had to do in Cargo.toml to make this compile, which should go away when released

[github-actions]

All checks have completed

❌ Failed Build / Build (pull_request)
❌ Failed Check / Check (docs-check) (pull_request)
❌ Failed Check / Check (gen-check) (pull_request)
❌ Failed Check / Check (lint-all) (pull_request)
❌ Failed Check / Check (mod-tidy-check) (pull_request)
❌ Failed Test / Cache Dependencies (pull_request)
⏭️ Skipped Test / Test (${{ matrix.name }}) (pull_request)

[rvagg]

Oh, and this currently doesn't get past the actors call for NI-PoRep WithRealProofs, it fails with:

sector proof size 14164 exceeds max 192

I believe we should have a single groth16 snark proof that rolls up the new PoRep for the batch (single) case so I that should be 192 as far as I understand. But somehow we're getting this massive proof for a single sector.

Thread about this in Slack @ https://filecoinproject.slack.com/archives/C06EZQ2DTD0/p1716637350693809

[rvagg]

FWIW this now passes if I comment out this line in actors: https://github.com/filecoin-project/builtin-actors/blob/308f0c06d01ae2fdc017fbe022a077bb972c3991/actors/miner/src/lib.rs#L2004

Discussion still ongoing regarding what the expected sizes are for proofs, but that check probably just needs to be changed to account for single NI-PoRep proof sizes which are much bigger than plain proofs.

[rvagg]

Added support for NI-PoRep in lotus-bench so I can generate --non-interactive proofs and do some testing.

I did break SealProofTypeFromSectorSize in chain/actors/builtin/miner though, it currently has a synthetic bool argument, I changed it to:

type SealProofVariant int

const (
	SealProofVariant_Standard SealProofVariant = iota
	SealProofVariant_Synthetic
	SealProofVariant_NonInteractive
)

// ...

func SealProofTypeFromSectorSize(ssize abi.SectorSize, nv network.Version, variant SealProofVariant) (abi.RegisteredSealProof, error) {
// ...
}

And it does slightly more rigorous checking on what's possible—the variant must be valid for the network version.

[rvagg]

Status of this is that (for the 3rd time) I'm going to drop this branch and push a new one soon. I've started again locally on top of master because most of the churn that went into this branch has been merged in the UnmanagedMiner work and we have the markets and client removal stuff on top of that so there's a huge diff and it's almost impossible to reconcile now. Plus, NI-PoRep will go into a new, separate itest now that we have pulled all the shared UnmanagedMiner stuff into the testkit.

[rvagg]

Replaced by #12105