[v6.x backport] doc,assert: document stackStartFunction in fail

[refack]

Refs: #13862
Was asked to port to v8.x but I ported to v6.x by mistake, so here it is if you want it 🤷‍♂️

/cc @nodejs/lts

Checklist

• make -j4 test (UNIX), or vcbuild test (Windows) passes
• tests and/or benchmarks are included
• documentation is changed or added
• commit message follows commit guidelines

Affected core subsystem(s)

assert

[refack]

CI: https://ci.nodejs.org/job/node-test-pull-request/9298/

[refack]

Fixed CI: https://ci.nodejs.org/job/node-test-commit/11311/

[refack]

• arm-fanned seems infra related, rerunning - https://ci.nodejs.org/job/node-test-commit-arm-fanned/10149/

• Windows fail seems unrelated but consistent across variants - https://ci.nodejs.org/job/node-test-binary-windows/10026/

not ok 179 parallel/test-net-listen-port-option --- duration_ms: 60.135 severity: fail stack: |- timeout ... 

• macos fail is a new flake [v6.x] test: investigate test-fs-read-buffer-tostring-fail #14430 - https://ci.nodejs.org/job/node-test-commit-osx/11278/nodes=osx1010/

[sam-github]

@refack rebase onto v6.x-master and not ok 179 parallel/test-net-listen-port-option will dissappear, at least.

I'm confused by the PR title "document ...", but it implements, as well as documents?

Adding yet another argument to assert.fail() seems reasonably safe, any reason we shouldn't backport this?

[sam-github]

Staring slightly harder, this seems to be a couple changes squashed into one commit: both a refactor of lib/assert.js, and the documentation of a pre-existing API?

@refack need some clarification here, please.

[BridgeAR]

@sam-github it is indeed a combination of multiple commits and if I read the change correct it also changes assert.fail a tiny bit. It also includes e.g. #12293

[refack]

@refack need some clarification here, please.

Original the request was to backport #13862 to v8.x. Out of haste I ported it to v6.x by mistake. To smooth port I dragged in #13974 as well.
It was major in master/v8.x but here AFAICT it's minor.

Anyway if it looks iffy feel free to reject it

[MylesBorins]

Decided not to move forward

[refack]

Now that #15479 (the backport of #12293) landed. This has become completely minor.

[refack]

CI: https://ci.nodejs.org/job/node-test-commit-linuxone/9226/ ✔️

[MylesBorins]

@refack the lts working group decided to not land this independent of how hard it was to backport. We gate which minor updates we land on the branch. Closing again, feel free to continue discussion though.

[refack]

As far as I remember (from listening to the discussion) the decision was based on this PR affecting the output of assert.fail. That part was just backported by #15479.
So now this PR is almost completely documentation and refactoring.

The only change is improving the default error for no args assert.fail():

diff --git a/test/parallel/test-assert-fail.js b/test/parallel/test-assert-fail.js index a64cfdb3abb..ebd962be751 100644 assert.throws( () =>{assert.fail()}, - /^AssertionError: undefined undefined undefined$/+ AssertionError,+ 'Failed' );

So I'd like this to be reconsidered.

[gibfahn]

As far as I remember (from listening to the discussion) the decision was based on this PR affecting the output of assert.fail. That part was just backported by #15479.

So the part that we thought was a blocker we've actually already backported? Seems a bit silly not to backport this now then.

+1 to backporting.