fix(nextcloud): give apache a real tmpfs for /var/run

Apache's own pidfile write there depends on an implicit overlay copy-up
succeeding, which in turn depends on the container's rootless UID mapping
matching whatever the image layer was extracted under. Found failing with
"(75) Value too large for defined data type" on a host whose subuid range
had drifted from the fleet-standard 100000 base -- nextcloud had been
running fine for days before that, since the implicit copy-up happened to
work under the old mapping by coincidence, not by design.

Matches indeedhub's identical --tmpfs /run precedent for the same class
of baked web-server scratch directory.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>

apps/nextcloud/manifest.yml

@@ -39,6 +39,16 @@ app:
       source: /var/lib/archipelago/nextcloud
       target: /var/www/html
       options: [rw]
+    # Writable scratch apache's own pidfile needs (matches indeedhub's
+    # identical --tmpfs /run precedent). Without an explicit tmpfs here,
+    # apache relies on an implicit overlay copy-up for /var/run/apache2,
+    # which depends on the container's rootless UID mapping lining up with
+    # whatever that layer was extracted under -- found failing with
+    # "(75) Value too large for defined data type" on a host whose subuid
+    # range had drifted from the fleet-standard 100000 base (2026-10-10).
+    - type: tmpfs
+      target: /var/run
+      options: [rw, nosuid, nodev, size=16m]
 
   environment: []