Add URL testing as CI step

In order to ensure we are notified of broken URLs, add testing of external
URLs as a step in the CI. It is added as `allow_failure` because it would
be inapproprate to stop merges because an unrelated link has broken.

Additionally, only test when CI is run on master branch to save repeatedly
testing the URLs each time a merge request or branch is updated.

Signed-off-by: Martyn Welch <martyn.welch@collabora.com>

.gitlab-ci.yml

@@ -2,9 +2,27 @@ variables:
   GIT_SUBMODULE_STRATEGY: recursive
 
 stages:
+  - test
   - generate-html
   - generate-pdf
 
+test-urls:
+  image: debian:buster-slim
+  stage: test
+  tags:
+    - lightweight
+  before_script:
+  - apt update && apt install -y --no-install-recommends
+      python3
+      python3-markdown
+      python3-requests
+  script:
+  - scripts/test_urls.py content/
+  # We are running this for all commits for all URLs on the website, so it
+  # wouldn't be fair to block a merge due to an unrelated URL suddenly failing.
+  # Treat this as a warning that the site has degraded.
+  allow_failure: true
+
 test-html:
   image:
     name: klakegg/hugo:latest