Align heading-id slugs with jgm/djot#393; pluggable ASCII-folding extension

[dereuromark]

Aligns auto-generated heading ids with the settled jgm/djot#393 rule, and adds ASCII
folding back as a pluggable, opt-in extension for sites that want maximum-portability
anchors.

Default change (spec-safe, per #393)

The identifier is formed by replacing each maximal run of non-alphanumeric ASCII with
a single - and trimming. Letter case and all non-ASCII characters are preserved.
This drops two djot-php-specific behaviors that diverged from the spec:

• always-on ASCII transliteration (Über->Uber) - now opt-in
• the _ exception - _ is non-alphanumeric ASCII, so it is replaced (#393 removed
  the per-character exceptions, #391)

input	before	after (#393)
Über café	Uber-cafe	Über-café
under_score	under_score	under-score
日本語	ri-ben-yu	日本語
a--b	a-b	a-b

A leading-digit result is prefixed s- (9 lives -> s-9-lives) for CSS-selector safety
(querySelector('#9-x') throws); orthogonal to #393, which governs punctuation only. This
PR also unifies that prefix from the old h- to s-, matching the existing empty-heading
s-N fallback. Empty results still fall back to s-N.

Opt-in: AsciiHeadingIdsExtension

ASCII folding is no longer a baked-in flag - it is a pluggable id transform. The core
HeadingIdTracker produces the pure #393 id and accepts an optional
Closure(string): string; AsciiHeadingIdsExtension sets that transform. It is wired to
both the renderer's tracker and the parser's heading-reference resolution, so
<section id> and implicit [Heading][] link targets stay in parity. The transform runs
over the spec id and is re-slugged, so a transform that reintroduces separators (e.g. CJK
romanization) still yields a clean id.

Retaining old behavior / web-safe ASCII URL ids (migration)

This is a breaking change to default heading-id output. To get behavior close to the
previous default - and ASCII-only, percent-encode-free, selector-safe anchors for web
publishing - add the extension:

$converter = new DjotConverter();
$converter->addExtension(new AsciiHeadingIdsExtension()); // Über café -> uber-cafe

Notes on backward compatibility:

• The extension restores ASCII folding, but the punctuation rule still follows #393
  (_ -> -) and the prefix is now s- (not h-), so a handful of ids differ even with
  the extension on. There is no flag that reproduces the old ids byte-for-byte - that is
  the deliberate spec-alignment.
• To pin a specific anchor regardless of the algorithm, set an explicit id on the heading
  (## Title {#my-id}) - explicit ids always win and are unchanged.
• Unicode ids in the new default are valid HTML5 and resolve in browsers (URL fragments
  percent-encode); ASCII is only needed for maximum portability, which the extension
  provides.

Notes

• The official djot test suite is unchanged (its id cases are simple ASCII,
  case-preserved). Internal tests that pinned the old transliterating behavior were
  updated to the #393 output; the extension and the pluggable transform have their own
  tests.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 91.86%. Comparing base (fe05794) to head (23ef085).

Additional details and impacted files

@@            Coverage Diff            @@
##             master     #224   +/-   ##
=========================================
  Coverage     91.85%   91.86%           
- Complexity     3497     3504    +7     
=========================================
  Files           104      105    +1     
  Lines          9921     9929    +8     
=========================================
+ Hits           9113     9121    +8     
  Misses          808      808           

☔ 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.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.