dashboards: Make the regex much more specific

To make sure that we're not being too greedy with the regex
replacement of the data source to use for each dashboard
that's uploaded, we make it much more specific using the
following:

- literal boundaries for " on either side of the match
- non-capturing optional group matches for the ${} bits
- the case-sensitive literal match for DS_ which is
  always in every exported dashboard representing the
  data source used
- a one-or-more case-sensitive match for the part that
  follows the underscore, with no characters allowed
  except alphabetical

This should prevent mistaking any capitalised labels or
anything else which may be present in the dashboard.

tasks/dashboards.yml

@@ -25,8 +25,8 @@
   become: no
   replace:
     dest: "/tmp/dashboards/{{ item.dashboard_id }}.json"
-    regexp: "[$].DS_.*}"
-    replace: "{{ item.datasource }}"
+    regexp: '"(?:\${)?DS_[A-Z]+(?:})?"'
+    replace: '"{{ item.datasource }}"'
   delegate_to: localhost
   run_once: yes
   with_items: "{{ grafana_dashboards }}"