Materialize responsive layout companion blocks - #1301

Tested this branch (5f812b0) paired with Automattic/blocks-engine#1191 (ae9716ef) on a real 28-page Wix store artifact (102 MB) through Studio's create --from flow (Automattic/studio#3952).

Result: all 28 pages compile, then materialization is rejected by two policy gates before anything renders.

1. SSI quality gate (fail_on_quality: true, which Studio sets): 2 unsupported fallbacks, both the site's contact form (desktop and mobile variants) — interactive_form / form:form_requires_runtime:server_or_client_form_handler, island type provider_materializable_form, repair class materialize_form_provider. The gate runs at plan_artifact, before the dependency lifecycle that could materialize the form, so materialize_dependencies: true and seed_entities: true do not help.
2. Blocks Engine editability policy (enforcement: required): max_nesting_depth 26, maximum 20, on 10 pages (36 failures). Wix wraps <main> in 21+ divs.

The plan-only route (wp static-site-importer materialize-wordpress-site-plan on the composed wordpress-site-plan/v2) is rejected with the same editability failure, plus 278 author_layout_topology_changed and 110 html_semantic_parity_navigation_menu_missing diagnostics. The composed plan holds 2 custom/responsive-layout, 54 custom/responsive-media, 1,614 custom/layout-shell, and 58 core/html blocks.

For comparison, Blocks Engine trunk 4c543747 + SSI main 6184b18 on the same artifact: the gates pass (the form sits inside the swallowed <main>), 32 pages materialize, and the frontend renders header and footer only (#1255 / blocks-engine#1143).

Size and time, this branch vs trunk (same artifact, same machine; both runs overlapped, so timings are inflated): shared plan 231 MB vs 39 MB; prepare_pages 354 s vs 174 s; compile_pages 728 s vs 427 s; compose 147 s both; composed result 842 MB vs 835 MB.

So on a Wix site this branch surfaces the real content, but the two gates stop it before a page renders. Happy to share the artifact and run directory privately.

AI assistance: Claude Code (Claude Fable 5) ran the comparison and drafted this comment. Aagam Shah directed the work and is responsible for it.