[jhaygood86]

Part 2 of a CSS 2.1 test-parity sweep against PeachPDF's Integration test suite. Stacked on #264 (float/clear/positioning) and #263 - the diff below includes both until they merge. Only the last 4 commits (counter engine, list-marker fixes, one stale-test fix) are this PR's own.

Summary

Scope: chapter 12 (generated content, automatic numbering, lists) - previously zero behavioral coverage in either test project (only content/list-style CSS-OM parsing).

• counter-reset/counter-increment: had zero effect anywhere - content: counter() was explicitly out of scope per CssContentEngine's own doc comment, and there was no counter-value tracking in Core at all. Implemented CssCounterEngine: a forward-order tree walk threading named counter values through document order (parent→first-child, sibling→sibling), covering ordinary sequential/nested counter usage. Deliberately scoped down from PeachPDF's full §12.4 algorithm - reversed() and counter-set aren't implemented; see the class's own remarks for the exact boundary. Ported PeachPDF's CssCounterEngineTests.cs (pure counter-style formatting, 17 tests) and the counter-related subset of CssContentEngineTests.cs, plus two new integration tests proving the document-order threading itself (sequential siblings, nested-ancestor visibility).
• List markers: <li value="N"> had no effect (only <ol start>/reversed were consulted) - fixed. Also found and fixed list-style-type:square rendering "♠" (a card-suit glyph, not a square) while auditing this code.
• Stale-Ignore audit: one existing [Ignore]d test claimed "this fork has no pseudo-element support at all" - false (::before clearly already worked, per the passing test right above it in the same file). Un-ignoring it surfaced a real bug in the test itself (a FirstOrDefault predicate ambiguous between the real text node and the ::after box), not the engine - fixed and passing.
• PeachPDF gap filed: jhaygood86/PeachPDF#863 - the quotes property (and open-quote/close-quote nesting depth per CSS 2.1 §12.2/§15.3) has no test coverage, and doesn't appear to be implemented there either.

Test plan

• HtmlRenderer.Test full suite: 2449 passed, 0 failed
• HtmlRenderer.IntegrationTest full suite: 374 passed, 0 failed
• HtmlRenderer.PdfSharp.Test full suite: 39 passed, 0 failed
• Full solution builds clean across net8.0, net8.0-windows, netstandard2.0, net462
• Re-approved the Bullets.png baseline for the corrected square-marker glyph - reviewed the diff image, confirmed it's the intended fix

[eXpl0it3r]

Rebased onto the updated #264 and fixed the counter scoping on top:

• A counter-reset on a nested element leaked out of its parent, so with <ol><li>A<ol>…3 items…</ol></li><li>B</li></ol> the outer list continued at 4. CssCounterEngine now keeps a stack of instances per counter name, a reset is in scope for the box's following siblings and their descendants, and goes away once the parent ends (§12.4.1). A second reset among the same siblings replaces the instance instead of nesting.
• Elements with display:none don't reset or increment counters anymore (§12.4.3).
• A counter that isn't in scope reads as 0 instead of 1, and incrementing one behaves as if the box reset it to 0.

Since the scoping is a proper stack now, counters() should be quite easy to add on top of it at some point. The three cases have tests in CssContentEngineIntegrationTests.

[eXpl0it3r]

Thank you!