core(entity-classification): classify unknown urls as "unattributable"

[alexnj]

Fixes #14996.

Previously, chrome:// and chrome-extension:// protocol URLs, as well as URLs that aren't part of devtoolsLogs would fail the audit. This change classifies all unknown as third party and groups them under "Unattributable".

With the axe chrome extension that @adamraine faced an error in #14996, the report now would look like:

[brendankenny]

The chrome:// and chrome-extension:// URLs part of this feels a bit weird, since they aren't unattributable, we know exactly what they are.

Not sure if chrome:// URLs are really an issue unless you're actually testing a chrome:// page, but should there be two changes here?

• chrome-extension:// URLs attributed to Host Chrome Extension(s) or whatever
• otherwise unknown URLs no longer throw an error and are put under Unattributable?

[alexnj]

@brendankenny We still recognize chrome:// pages as a supported URL-type, so we don't want the audits to fail if they are trying to classify those URLs.

I'm not sure if we are adding further value by classifying chrome extensions as a separate entity of their own, and we'll have to group all chrome extensions together anyway, so they're likely best left grouped into unattributable entities.

[brendankenny]

We still recognize chrome:// pages as a supported URL-type, so we don't want the audits to fail if they are trying to classify those URLs

Right, that's what I meant by they're not an issue unless you're actually testing a chrome:// page since AFAIK you can't load resources from there otherwise. Maybe we should treat all chrome:// URLs as the same entity?

I'm not sure if we are adding further value by classifying chrome extensions as a separate entity of their own, and we'll have to group all chrome extensions together anyway, so they're likely best left grouped into unattributable entities.

It makes sense to not make unattributable URLs a fatal error that breaks audits, but that error was added because we believe LH should be able to know where all URLs came from, and examples to the contrary are corner cases that need to be fixed. Here's a corner case :) We know what they are and we should say so.

Grouping them all together seems fine. It's the dev's extensions, so as long as we make that evident (Chrome extension(s) on test machine?), that's a clear handoff of responsibility at that point.

There also doesn't seem to be any downside? An extra string, but makes the classification clearer and more authoritative.

[alexnj]

Okay let's merge this in so we can fix the audit failure, and I'll follow up with an issue to classify Chrome extensions separately. @brendankenny wdyt?

[brendankenny]

SGTM!