feat(modules): the client registry, window.__rg and the chunk's script injection

Phase 2, PR 7 of docs/website/MODULE_SYSTEM.md 2.7 — the client half's
delivery. A module's prebuilt chunk is served, injected, handed core's React
and its UI kit, and its routes are rendered by App.jsx. The registry is empty
on a bare core, so nothing an operator can see changes.

Client:
  - modules/registry.js — registerRoutes/registerNav/registerFeatureProvider,
    with the URL namespace written by core, never by the module
  - modules/shared.js — window.__rg: React, react-dom/client, react-router-dom,
    react/jsx-runtime, the registry, the seven-member UI kit and the request
    primitive, frozen
  - App.jsx reads routesFor for all three areas; nav consumption is PR 8
  - main.jsx publishes the global, then mounts on DOMContentLoaded

Server:
  - the loader validates client.entry and publishes clientChunks() and
    clientEntryUrls(); an entry in the module root is rejected, because the
    directory it sits in is what gets served
  - app.js mounts each chunk at /modules/<id>/ behind the module's state guard
    with no-cache; anything else under /modules is a 404, not the SPA shell
  - htmlShell injects the tag before </body>, so core's bundle runs first
    wherever a bundler puts it

Found by loading a real chunk in a browser, and fixed here: core mounted before
any module chunk had evaluated, because document.readyState during a deferred
script is 'interactive', not 'loading'. Every test passed against that build.
The smoke is written down in MODULE_API.md 7.7.

933 server tests (+23), 123 client tests (+14). routes.manifest.json unchanged
at 230 routes; the OpenAPI spec regenerates byte-identical.

Co-Authored-By: Claude <noreply@anthropic.com>

client/src/main.jsx

@@ -2,12 +2,54 @@ import React from 'react'
 import { createRoot } from 'react-dom/client'
 import { BrowserRouter } from 'react-router-dom'
 import App from './App.jsx'
+import { publishSharedDependencies } from './modules/shared.js'
 import './styles/theme.css'
 
-createRoot(document.getElementById('root')).render(
-  <React.StrictMode>
-    <BrowserRouter>
-      <App />
-    </BrowserRouter>
-  </React.StrictMode>,
-)
+// Publish window.__rg BEFORE rendering and before any module chunk evaluates.
+// Installed modules arrive as `<script type="module" src="/modules/<id>/…">`
+// tags the server injects into the shell (server/src/utils/htmlShell.js), placed
+// after this bundle's own tag; module scripts execute in document order, so they
+// resolve their externals against the global this call sets up
+// (docs/website/MODULE_API.md §3.2).
+publishSharedDependencies()
+
+// Render on DOMContentLoaded rather than immediately, and that is the one line
+// of core's boot the module system changes.
+//
+// Deferred scripts — which every `type="module"` script is — execute in document
+// order and ALL of them finish before DOMContentLoaded fires. Waiting for that
+// event is therefore the guarantee that every installed module has registered
+// its routes before React reads the registry: no loading state, no re-render,
+// and no ordering race between core's bundle and a module's. A module chunk that
+// 404s or throws does not hold the event back, so a broken module costs its own
+// pages and not the site.
+//
+// The readyState check below is `'complete'`, and it is not the obvious
+// `'loading'`. A DEFERRED script — which every `type="module"` script is — runs
+// after the document has been parsed, so by the time this line executes
+// readyState is already `'interactive'`; DOMContentLoaded has NOT fired yet and
+// still comes after every deferred script. Testing for `'loading'` therefore
+// mounts immediately, before any module chunk has evaluated, and a module's
+// routes are missing from the very first render — which looks exactly like a
+// module that failed to load: its URL falls through to core's catch-all and
+// redirects home. Found by loading a real chunk in a browser; no unit test in
+// this repo can see it.
+//
+// `'complete'` is only reached after `load`, which is strictly later than any
+// static deferred script, so this branch is the genuine "the event has already
+// been and gone" case and not a wrong guess about our own timing.
+function mount() {
+  createRoot(document.getElementById('root')).render(
+    <React.StrictMode>
+      <BrowserRouter>
+        <App />
+      </BrowserRouter>
+    </React.StrictMode>,
+  )
+}
+
+if (document.readyState === 'complete') {
+  mount()
+} else {
+  document.addEventListener('DOMContentLoaded', mount, { once: true })
+}