module: replace default paths in require.resolve()

[cjihrig]

Prior to this commit, the default search paths would be included in the require.resolve() process, even if user specified paths were provided. This commit causes the default paths to be
omitted by using a fake parent module.

Refs: #5963 (comment)

Checklist

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

Affected core subsystem(s)

module

[refack]

Code LGTM, and IMHO semantic makes sense.

[refack]

Could you add .js to the last arg?
(much debate about extensionless requires related to ESM, and future ESM CJS interop)

[refack]

This is obviously a bug fix, when compared to the docs

node/doc/api/modules.md

Lines 611 to 615 in 707cd3f

	* `options` {Object}
	* `paths` {Array} Paths to resolve module location from. If present, these
	paths are used instead of the default resolution paths. Note that each of
	these paths is used as a starting point for the module resolution algorithm,
	meaning that the `node_modules` hierarchy is checked from this location.

But since CJS resolution is fragile, how could we minimize possible breakage besides running CitGM?

CitGM: https://ci.nodejs.org/view/Node.js-citgm/job/citgm-smoker/1084/

[cjihrig]

But since CJS resolution is fragile, how could we minimize possible breakage besides running CitGM?

I doubt the ecosystem usage is too significant since this feature has existed for less than a month.

[jure]

This is lengthier +1, but hopefully carrying a bit more information. We've looked at 8.9's require.resolve, as the docs seemed to fit our need for it (we currently use require-relative for it). We tried it and it didn't work, we spend a good deal of time making minimal test cases, looking at the PR that introduced the feature and pondering why it doesn't work like intended (and how come nobody has complained, either in the PR or after!). We settled on keeping require-relative, which basically does a similar thing as this PR proposes (creates a new Module).
Point being, we're happy to see this PR! One of our devs has compiled node with it and confirms that it now works as we expected it to initially. 👍

[cjihrig]

CI: https://ci.nodejs.org/job/node-test-commit/14204/

[MylesBorins]

letting it bake for a bit

thoughts on how long @cjihrig ?

[cjihrig]

Until the temperature in the middle reaches 350 degrees.

It looks like #16397 is already in Node 8, but it's a bit broken, so I'd say if you don't take it this time, then next time.

[gibfahn]

so I'd say if you don't take it this time, then next time.

Sounds reasonable, landed.

[MylesBorins]

setting don't land for v6.x as #16397 didn't land on that release line