Decompose animated marketing sections into native blocks instead of core/html app-shell fallback

[chubes4]

Problem

Ordinary marketing <section>s were preserved wholesale as a single core/html block whenever animation/counter scripts targeted elements inside them. shouldPreserveRuntimeAppShell() counted presentation-only selectors (.reveal, [data-count], animate/motion/scroll classes) fed by ArtifactCompiler::runtimeDomSelectors() as runtime app-shell evidence, so any section with scroll-reveal animation looked like an interactive app and fell back — the single largest fallback family in the Static Site Importer fixture matrix (12 findings across 4 fixtures on the complexity-1 lane).

Fix

• Filter presentation-only animation/counter selectors in isRuntimeDomTarget() and data-attribute preservation — generalizing the existing data-animation filter to class/id/data selectors. Structural decision only; no fixture- or class-name-specific matching beyond generic animation vocabulary.
• data-count support in dynamicTextContent() so counter text stays editable.
• New parity fixture: html-section-runtime-animation-native-decomposition.json.

Verification

• Full suite green: composer test (canonical + 203 parity fixtures + packaging).
• Static Site Importer fixture matrix (complexity-1 lane, 7 fixtures, real WP sandbox with editor-side wp.blocks.validateBlock):
  • fallback_block:core_html_block:section family: 12 → 0
  • Total blocks 917 → 1,142 (+225 real editable blocks from decomposed sections)
  • Editor-valid block rate stays 1.0; 7/7 fixtures pass the gate
  • Expected cascade: inner SVGs previously buried in section fallbacks now surface individually (svg family 8 → 19), tracked as follow-up work

AI assistance

• AI assistance: Yes
• Tool(s): Claude Opus 4.5 (claude-fable-5) via opencode, orchestrated agent workflow
• Used for: Root-cause investigation, implementation, parity fixture, and fixture-matrix verification runs. Reviewed and verified by Chris Huber via the SSI fixture matrix quality gates.