Fix no-descending-specificity performance

[romainmenke]

Which issue, if any, is this issue related to?

See #6869

Is there anything in the PR that needs further explanation?

This rule warned multiple times for the same selector.
It would emit one warning for each selector with lower specificity that preceded the current selector.

I don't think this is very useful as it just causes extra noise.

With this change this rule will only emit a single warning for each selector.
It will always reference the earliest other selector with lower specificity.

All these extra warnings came at a cost.
fewer warnings -> faster execution.

I've also added extra tests to cover this scenario.

---

before :

Warnings: 407860
Mean: 1452.8959879999998 ms
Deviation: 85.89908815057446 ms

after :

Warnings: 7117
Mean: 367.08543172727275 ms
Deviation: 14.07386884681592 ms

---

The excessive number of warnings is the result of the duplicate CSS in the test source.

Ideally we would have a large, but well formed and realistic benchmark source.

[changeset-bot]

🦋 Changeset detected

Latest commit: b81d660

The changes in this PR will be included in the next version bump.

This PR includes changesets to release 1 package

Name	Type
stylelint	Patch

Not sure what this means? Click here to learn what changesets are.

Click here if you're a maintainer who wants to add another changeset to this PR

[jeddy3]

Nice, this was the slowest rule I encountered while taking a pass at #6869. So, it's great to see this improvement.

(I saw your performance-related pull request upstream in PostCSS too. Excellent stuff.)

[ybiquitous]

Great improvement! 👏🏼

Rule#selectors property actually has some calculation. Honestly, I feel it's hard to notice it... 😓
https://github.com/postcss/postcss/blob/8.4.24/lib/rule.js#L13-L15

[ybiquitous]

Sorry for the conflict in the test file due to #7024. Could you resolve the conflict, please? 🙏🏼