tools: update to mdast-util-to-hast v3.0.2

[rubys]

See syntax-tree/mdast-util-to-hast#21

Fixes #22065

Note: I updated all of the tools/doc dependencies, not just this one,
and removed the previous workaround that was in place until this change
landed.

Checklist

• make -j4 test (UNIX), or vcbuild test (Windows) passes
• commit message follows commit guidelines

[nodejs-github-bot]

@rubys build started: https://ci.nodejs.org/blue/organizations/jenkins/node-test-pull-request-lite-pipeline/detail/node-test-pull-request-lite-pipeline/436/pipeline

[vsemozhetbyt]

There is good news.

This PR fix two known regressions and one previously unknown regression:

1. Regression with Markdown escapes described in doc: header escaping regression #22065

2. Regression with HTML entities described in doc: fix header escaping regression (Issue #22065) #22084 (comment)

3. Unknown regression with Markdown backticks rendered in headings instead of code tags (nightly example).

[vsemozhetbyt]

There is bad news.

This PR introduces one new regression. If a heading has two consecutive optional parameters in square brackets, they are not rendered as links, but some preprocess cleaning is done as if they are links:

1. spaces between them are deleted;
2. all text in the second pair of brackets is lowercased.

Compare these examples before and after this PR:

[vsemozhetbyt]

And there is good bad news)

Comparing the diff helps to find out one issue that is differently exposed before and after this PR.

1. Before this PR. If a parameter type signature is a 2-dimensional array with a [][] part, this part is rendered as an invisible empty link (nightly example, see records parameter) and the type is not properly shown and linkified.

2. After this PR. All the type is rendered wrongly verbatim without proper linkification:

Compare the doc from the old toolchain: https://nodejs.org/docs/latest-v8.x/api/dns.html#dns_dns_resolvetxt_hostname_callback

One of the concerned code fragments:

node/tools/doc/type-parser.js

Lines 138 to 141 in 28c70ac

	// To support type[], type[][] etc., we store the full string
	// and use the bracket-less version to lookup the type URL.
	const typeTextFull = typeText;
	typeText = typeText.replace(arrayPart, '');

[rubys]

I know what's going on. I'll try to see if i can create another patch for mdast-util-to-hast.

[rubys]

Three pull requests created to solve (most) of the problems noted above:

• Add reference field to LinkReference, ImageReference syntax-tree/mdast#23
• Add reference property on full references when different from identif… remarkjs/remark#346
• Prefer reference over identifier when reverting an unresolved reference. syntax-tree/mdast-util-to-hast#22

The 2-dimensional array appears to be a separate problem in that remark, et. al., appears to be doing the right thing. Once the above fixes land, I'll look into fixing that problem too.

[Trott]

Should this update the corresponding package.json file and not just the package-lock.json? I mean, I guess that's not required, but if we want to be sure we are bumping a dependency up to a certain minimum version, that should be reflected in the package.json?

[rubys]

@Trott will do... once the above three patches are resolved. But first, there are yaks that need shaving: syntax-tree/mdast#23 (comment)

[jasnell]

Ping @rubys ... what's the status on this one?

[trivikr]

The PRs listed in #22140 (comment) need a follow-up
Ping @rubys

[Trott]

@rubys Looks like all upstream yaks have been shaved and this can be rebased, run through CI, and hopefully landed?

[Trott]

Took the liberty of rebasing. Hope that's OK. Still have to install the upstream dependencies that have been updated and run through CI?

[BridgeAR]

CI https://ci.nodejs.org/job/node-test-pull-request/21382/

[refack]

CI: https://ci.nodejs.org/job/node-test-pull-request/21385/