Replace docker-compose restart hack with shell handler hack

The docker-compose integration would start 2 of the same container, which does bad things to things like databases!

ansible/roles/traefik/handlers/main.yml

@@ -0,0 +1,4 @@
+- name: restart traefik
+  shell:
+    chdir: /opt/traefik
+    cmd: "{{ docker_update_command }}"

ansible/roles/traefik/tasks/main.yml

@@ -20,7 +20,7 @@
     mode: "{{ docker_compose_file_mask }}"
     owner: "{{ docker_user.name }}"
     validate: /usr/bin/docker-compose -f %s config
-  register: compose_file
+  notify: restart traefik
   become: true
 
 - name: Install config
@@ -29,7 +29,7 @@
     dest: /opt/traefik/traefik/traefik.yml
     mode: "{{ docker_compose_file_mask }}"
     owner: "{{ docker_user.name }}"
-  register: config_file
+  notify: restart traefik
   become: true
 
 - name: Install file provider
@@ -38,17 +38,5 @@
     dest: /opt/traefik/traefik/file-provider.yml
     mode: "{{ docker_compose_file_mask }}"
     owner: "{{ docker_user.name }}"
-  register: file_provider
+  notify: restart traefik
   become: true
-
-- name: Cycle container
-  docker_compose:
-    project_src: /opt/traefik
-    pull: true
-    remove_orphans: true
-    remove_volumes: true
-    state: "{{ item }}"
-  when: compose_file.changed or config_file.changed or file_provider.changed
-  loop:
-    - absent
-    - present