feat(modules): the three de-entanglement registries, with core as the registrant

Phase 2 PR 4 of docs/website/MODULE_SYSTEM.md §2.7. Adds server/src/modules/registries.js
and moves core's own notification streams, announce leg and users-detail routes
behind it, so the three seams §1.8 and §1.9 named are exercised on every boot
before any module depends on them.

Registering is validate-then-commit per registrant: the loader stages what a
module claims and the second pass commits it, so a module that throws halfway
through register() — or fails checkDeclared after it — leaves nothing behind.
That is the registry-side twin of PR 2's second-pass mount rule.

Four decisions, all the recommended option:

- announce legs became a child table. `announce_job_legs` replaces the
  towncrier_*/discord_* column groups, so the leg set is data: core registers
  `discord`, module-uo will register `towncrier`, and a module cannot ALTER a
  core table to add its own. Backfill is guarded on information_schema (a
  SELECT of a dropped column is a parse error, not a runtime one) and the
  columns go with DROP COLUMN IF EXISTS. Verified against the live dev DB:
  three legacy jobs migrated faithfully, three replays, no duplicates.
- `mapEvent` dropped from registerNotificationStreams. §1.8 already inverts the
  push path so a module owns fromShardEvent and calls core's publish() with a
  stream id it resolved; a second mapping mechanism was a leftover. The public
  safety filter, the kinds it reads and the streams it protects now live in one
  file and move together.
- core registers through the same staging area a module uses, via an explicit
  registries.registerCore() in app.js before modules.load().
- core's six /admin/users/:id/shard/* paths now go through the
  `admin.users.detail` slot, and getUser moved back to admin.controller.js.

Found on the way, and the reason two build tools changed:

- scripts/routeManifest.js could not decode a parameterised mount. Its
  unwinder expected `(?:([^\/]+?))`; express 4.22 emits `(?:\/([^/]+?))` with
  the separator inside the group. The branch had never run. It threw rather
  than guessing, which is what it is for.
- swagger-autogen cannot follow a route into an extension slot — the slot's
  router is created by declareSlot() and filled later, so there is no literal
  mount for a static parse. Regenerating deleted 407 lines and printed
  `Swagger-autogen: Success`, the spike's exact failure (MODULE_API.md §7.4).
  swagger/slotSpecs.js generates a fragment per filled slot and re-roots it at
  the prefix the router actually hangs at in the live app — read from the
  express stack via routeManifest's own mountPath, so the manifest and the spec
  cannot disagree. swagger/mergeSpec.js is the merge helper core owes for
  module fragments anyway (§6.1a), proved here against core's own slot first.

884 tests pass (856 before). routes.manifest.json is unchanged at 229 routes.
The OpenAPI spec diff is two lines of intent: the retry endpoint's summary, and
its `leg` no longer being a fixed enum.

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

server/swagger/mergeSpec.js

@@ -0,0 +1,115 @@
+// ── Merging an OpenAPI fragment into a spec ────────────────────────────────
+//
+// The merge half of docs/website/MODULE_API.md §6.1's settled decision: routes
+// that reach the app through something swagger-autogen cannot statically follow
+// contribute a FRAGMENT, and core merges it.
+//
+// Two callers, one function, deliberately:
+//   • build time — swagger/slotSpecs.js, for core's own extension-slot routers
+//     (§1.9). Those are core routes that a static parse of app.js cannot see,
+//     because the slot's router is created by registries.declareSlot() and filled
+//     later. They belong in the committed `swagger-output.json`.
+//   • request time (Phase 2 PR 6+) — an installed module's `swagger-fragment.json`,
+//     merged over the committed spec for `/api/docs.json`.
+//
+// **Core always wins a key collision** (§6.1a). A fragment cannot redefine a path,
+// a tag or a schema core already declares; the collision is reported and the
+// fragment's version dropped. Merging is shallow-per-section — `paths`, `tags`
+// and `components.schemas` — because those are the only three sections a fragment
+// is allowed to carry, and a deeper merge would let a fragment reach into
+// `info`, `servers` or the security schemes.
+
+/**
+ * Merge `fragment` into `spec`, in place, with core winning every collision.
+ *
+ * @param {object} spec      the base spec — mutated
+ * @param {object} fragment  `{ paths?, tags?, components?: { schemas? } }`
+ * @param {string} source    who the fragment came from, for the collision message
+ * @returns {string[]} the collisions that were dropped (empty when clean)
+ */
+function mergeFragment(spec, fragment, source) {
+  const dropped = []
+
+  for (const [path, item] of Object.entries(fragment.paths || {})) {
+    if (spec.paths[path]) {
+      // Not a merge of the two path items: a fragment adding a METHOD to a core
+      // path is the same overreach as replacing it, and the extension-slot
+      // contract already says core owns the resource (§2.4).
+      dropped.push(`path ${path}`)
+      continue
+    }
+    spec.paths[path] = item
+  }
+
+  const tagNames = new Set((spec.tags || []).map((t) => t.name))
+  for (const tag of fragment.tags || []) {
+    if (tagNames.has(tag.name)) continue // same tag, not a collision worth reporting
+    spec.tags.push(tag)
+    tagNames.add(tag.name)
+  }
+
+  const schemas = (fragment.components && fragment.components.schemas) || {}
+  for (const [name, schema] of Object.entries(schemas)) {
+    if (spec.components.schemas[name]) {
+      dropped.push(`schema ${name}`)
+      continue
+    }
+    spec.components.schemas[name] = schema
+  }
+
+  if (dropped.length > 0) {
+    process.stderr.write(
+      `swagger: dropped ${dropped.length} colliding key(s) from ${source} — core wins: ${dropped.join(', ')}\n`,
+    )
+  }
+  return dropped
+}
+
+/**
+ * Re-root a fragment's paths under the prefix its router is actually mounted at.
+ *
+ * A fragment generated by pointing swagger-autogen at a router file alone has
+ * paths relative to that router (`/shard/accounts`), because nothing in the file
+ * says where it hangs. The prefix comes from the LIVE express stack rather than a
+ * table, so it cannot drift the way a hand-written mount list would.
+ *
+ * Express path params (`:id`) become OpenAPI's (`{id}`) — swagger-autogen already
+ * does that for the paths it generates, so the prefix has to match.
+ */
+function prefixPaths(fragment, prefix) {
+  const oas = prefix.replace(/:([A-Za-z0-9_]+)/g, '{$1}').replace(/\/+$/, '')
+  const outer = [...oas.matchAll(/\{([A-Za-z0-9_]+)\}/g)].map((m) => m[1])
+  const paths = {}
+  for (const [p, item] of Object.entries(fragment.paths || {})) {
+    paths[`${oas}${p}`] = orderParams(item, outer)
+  }
+  return { ...fragment, paths }
+}
+
+/**
+ * Put the prefix's own path parameters first, in prefix order.
+ *
+ * swagger-autogen orders parameters by where they appear in the path it saw, and
+ * the fragment's path is only the tail — so `/{id}/shard/link/{account}` comes
+ * out as (account, id) rather than (id, account). Re-rooting the path has to
+ * re-root the parameter order with it, or every slot route churns the committed
+ * spec by a reorder that means nothing.
+ */
+function orderParams(item, outer) {
+  for (const operation of Object.values(item)) {
+    const params = operation && operation.parameters
+    if (!Array.isArray(params)) continue
+    const rank = (p) => {
+      const i = outer.indexOf(p && p.name)
+      return i === -1 ? outer.length : i
+    }
+    // Stable: only the prefix params move, and only ahead of the rest.
+    operation.parameters = params
+      .map((p, i) => ({ p, i }))
+      .sort((a, b) => rank(a.p) - rank(b.p) || a.i - b.i)
+      .map(({ p }) => p)
+  }
+  return item
+}
+
+module.exports = { mergeFragment, prefixPaths }