LibWeb: Don't inflate margin-right for over-constrained blocks

[wakamex]

getComputedStyle().marginRight returns an inflated value for over-constrained blocks (e.g. 400px instead of 0px), disagreeing with Chrome, Firefox, and Safari.

When a block's width, margin-left, and margin-right are all explicitly set and don't fill the containing block, CSS 2.2 § 10.3.3 says to inflate margin-right to satisfy the constraint equation. Since a706d0e switched getComputedStyle() to return used values for margins, this inflated value now leaks to the web. All major browsers store the declared value instead — see CSSWG #2328.

The inflated value serves no layout purpose: position is determined by margin-left, width is already set, and siblings position independently.

Reproducer (paste in DevTools console on any page):

let d = document.createElement('div');
d.style.cssText = 'width:100px; margin:0px';
document.body.appendChild(d);
getComputedStyle(d).marginRight;
// Ladybird: "700px" — Chrome/Firefox/Safari: "0px"

First commit adds a failing test, second commit applies the fix and rebaselines 365 layout test expectations.

One WPT test flips from fail to pass in css/css-logical/inheritance.html. The assert_not_inherited('margin-inline-end', '0px', '10px') now passes because the parent's margin-inline-end no longer gets inflated as it's an over-constrained case.

One WPT ref test is a genuine regression, as the existing "correct" solution relied on inflation: floats-wrap-bfc-outside-001.xht.

Fixes #4729
Supersedes #6606

[ladybird-bot]

Hello!

One or more of the commit messages in this PR do not match the Ladybird code submission policy, please check the lint_commits CI job for more details on which commits were flagged and why.
Please do not close this PR and open another, instead modify your commit message(s) with git commit --amend and force push those changes to update this PR.

One outstanding issue

[gmta]

@wakamex multiple tests are failing. Please check the test results locally and see if they need updating.

[github-actions]

Your pull request has conflicts that need to be resolved before it can be reviewed and merged. Make sure to rebase your branch on top of the latest master.

[wakamex]

Fixed 2 of the 3 failing tests. 1 was a missing --rebalance due to rebase. 2 is a WPT test that flips from fail to pass. 3 is a real regression that relied on the inflated margin. I'll have to take a closer look at that one. latter 2 I documented in the PR body as such:

One WPT test flips from fail to pass in css/css-logical/inheritance.html. The assert_not_inherited('margin-inline-end', '0px', '10px') now passes because the parent's margin-inline-end no longer gets inflated as it's an over-constrained case.

One WPT ref test is a genuine regression, as the existing "correct" solution relied on inflation: floats-wrap-bfc-outside-001.xht.

also found a material change in the last --rebaseline that needs fixing (gist)