Make filter() narrow the match set instead of keeping elements that contain a match

[jakejackson1]

Fixes the containment bug in filter() — the one all three of the open selector PRs ran into
and left alone because fixing it means deliberately rebaselining an existing test.

Based on issue-62 (#69), which is based on issue-51 (#67). Merge order: #67 → #69 → this.
It needs QueryPath\Helpers\NodeMatcher, introduced in #67.

The bug

filter() ran a descendant search per node, so it kept any element that merely contained a
match rather than narrowing the set to the elements that match:

$html = '<div><ul><li class="x">a1</li><li>a2</li></ul><p class="x">p1</p><span>s1</span></div>';

html5qp($html)->find('ul, p, span')->filter('.x');
// before: ul, p     <- ul only contains .x
// after:  p         <- jQuery

That is what has() does, and has() keeps doing it — it is the migration path for anyone
relying on the old behaviour.

The author's comment

filter() carried this since 2009:

// Seems like this should be right... but it fails unit
// tests. Need to compare to jQuery.
// $query = new \QueryPath\CSS\DOMTraverser($tmp, TRUE, $m);

It fails exactly one test — DOMQueryTest::testFilter, which asserts the containment result:

$this->assertEquals(1, qp($file)->filter('li')->count());          // <root> contains li
$this->assertEquals(2, qp($file, 'inner')->filter('li')->count()); // <inner> contains li

Under jQuery semantics both are 0. The test is rebaselined, and two tests are added: one
pinning the filter()/has() difference so the migration path stays covered, one covering
set-level evaluation.

Filtering the set, not the node

Both filter() and children($selector) now filter their candidates in one pass instead of
one node at a time. A per-node pass cannot evaluate a selector describing a position within the
set — each node would be the only member of its own one-element set.

This also drops the per-node scope node, fixing the same :scope bug that #69 fixed in the
traversal methods:

html5qp($html)->find('div')->children(':scope')->count();
// before: 3 (every child)   after: 0

children() did not have the containment bug — it already used the correct traverser mode.
Only its scope handling was wrong.

Interaction with #66

#66 documents a known limitation: because children() and filter() built a traverser per
node, a positional pseudo-class saw a one-element set, so children('li:first') returned every
li child. Filtering set-at-a-time resolves it. Verified by merging #66 into this branch
locally — 382 tests pass, and:

$q->find('li')->filter(':first')  // a1        (was: every li)
$q->find('li')->filter(':eq(3)')  // b1
$q->find('li')->filter(':odd')    // a2,b1
$q->find('ul')->children('li:first') // x1

No change is needed in #66; this branch just removes the reason for its caveat.

Verification

• vendor/bin/phpunit — 346 tests, 1164 assertions, 0 failures (2 pre-existing create_function skips)
• composer run lint and composer run lint:min-php — clean
• Merged with jQuery positional pseudo-classes (:eq, :first, :lt, :gt, :odd, :even) index siblings instead of the match set; :eq(0) never matches #66: 382 tests, 0 failures

🤖 Generated with Claude Code

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 89.71%. Comparing base (6ccf341) to head (aa85469).

Additional details and impacted files

@@              Coverage Diff               @@
##             issue-62      #71      +/-   ##
==============================================
- Coverage       89.74%   89.71%   -0.03%     
  Complexity       1363     1363              
==============================================
  Files              27       27              
  Lines            3071     3064       -7     
==============================================
- Hits             2756     2749       -7     
  Misses            315      315              

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[jakejackson1]

Collapsed into #72, which carries all of this work unchanged — same tree, one branch. Closing to keep the review in one place; reopen if the split turns out to be preferable.