fix(web-app-pgadmin): flush the deploy handlers before the post-deploy step

sys-stk-full only notifies the deploy handler and compose_handlers_flush defaults to false, so the post-deploy include ran before the stack existed and the health wait burned 60 retries on a container nobody had created. The CI job shows the notify at 10:07:22 while the last swarm stack deploy handler ran at 09:45:03, and the failure reports no running task for the service. The flush was present in 0709de6fa1 and was lost when the merge replaced administrator.yml with manager_ops.yml; this restores it in the same position, after the configuration render so both notifies fold into one deploy.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

roles/web-app-pgadmin/tasks/main.yml

@@ -7,6 +7,9 @@
   include_tasks: configuration.yml
   when: PGADMIN_SERVER_MODE | bool
 
+- name: "🚿 Flush compose handlers for {{ application_id }}"
+  ansible.builtin.meta: flush_handlers
+
 - name: "🖥️ pgAdmin stack-host post-deploy (administrator realign)"
   include_tasks: manager_ops.yml
   when: IS_STACK_HOST | bool