feat: configurable external check failures

[asterite]

Description

Problem

Resolves #6733

Summary

Now when we run external checks, we compare the nargo test failure results against an expected set of failures (these go in .github/critical_libraries_status). If the failures don't match, CI fails. This way if we make a change in Noir that breaks some tests in a repository, we can temporarily ignore those failures by saying that they are supposed to fail. Because CI will do a diff of the failures, if the failure gets resolved on the external library, the diff will fail and we'll know we have to remove that known failure.

If a change in Noir completely breaks a repository, we can delete (or rename) the .expected.jsonl file from this repo to signal that we know it's failing (and we'll know if it gets fixed and we didn't put back that file).

Additional Context

• To make this PR pass I had to comment out checking against rollup-lib because it was either running out of memory or timing out.
• There's also noir_bigcurve which takes a lot of time to run, but passes.
• Let me know if we should document all of this in a README file. I don't know if we'd also want to also run external checks on every PR from now on.

Documentation

Check one:

• No documentation needed.
• Documentation included in this PR.
• [For Experimental Features] Documentation to be submitted in a separate PR.

PR Checklist

• I have tested the changes locally.
• I have formatted the changes with Prettier and/or cargo fmt on default settings.

[github-actions]

Peak Memory Sample

Program	Peak Memory
keccak256	78.15M
workspace	122.43M
regression_4709	423.26M
ram_blowup_regression	1.58G
private-kernel-tail	206.67M
private-kernel-reset	720.25M
private-kernel-inner	291.85M
parity-root	171.68M

[github-actions]

Compilation Sample

Program	Compilation Time	%
sha256_regression	0m1.453s	8%
regression_4709	0m0.830s	1%
ram_blowup_regression	0m14.712s	3%
rollup-base-public	3m42.834s	8%
rollup-base-private	3m1.545s	-2%
private-kernel-tail	0m1.174s	-5%
private-kernel-reset	0m8.231s	-5%
private-kernel-inner	0m2.237s	-3%
parity-root	0m0.843s	-27%
noir-contracts	3m0.625s	12%

[asterite]

Strange, rollup-lib always gets stuck, even though it works fine on my machine.

[vezenovm]

Strange, rollup-lib always gets stuck, even though it works fine on my machine.

I noticed the same thing when making the compilation report (#6750 (comment)). Have not had the chance to investigate yet, I'll look a bit deeper at why this is occurring.

EDIT: It looks like for compilation that we have reduced memory enough that compiling the rollup circuits does not mess with the github runner. Perhaps that is not yet the case for testing.

[asterite]

Perhaps that is not yet the case for testing

Good point. I tried running the tests with one thread and the maximum memory used was like 3 GB. But with multiple threads it reaches 27 GB on my machine. That probably killed by GitHub.

I'll try reducing the number of test threads just for that project.

[michaeljklein]

LGTM

Let me know if we should document all of this in a README file.

It's not blocking, but I'd like to see a bit of docs in a README, e.g. a minimal one linking to this PR

[TomAFrench]

I think we can disable it for all external checks tbh. Fixing any issues associated with it falls to the external dev.

[github-actions]

Changes to Brillig bytecode sizes

Generated at commit: 3bcee054b450a534950d3147150392e853820448, compared to commit: e341a3b8561d37180335df49213bcc070717384c

🧾 Summary (10% most significant diffs)

Program	Brillig opcodes (+/-)	%
sha256_regression	-3 ✅	-0.04%

---

Full diff report 👇

Program	Brillig opcodes (+/-)	%
sha256_regression	6,917 (-3)	-0.04%

[github-actions]

Changes to number of Brillig opcodes executed

Generated at commit: 3bcee054b450a534950d3147150392e853820448, compared to commit: e341a3b8561d37180335df49213bcc070717384c

🧾 Summary (10% most significant diffs)

Program	Brillig opcodes (+/-)	%
sha256_regression	+3 ❌	+0.00%

---

Full diff report 👇

Program	Brillig opcodes (+/-)	%
sha256_regression	118,707 (+3)	+0.00%