Fix gh-pages deploy race condition

Replace pull_request_target with pull_request to prevent
duplicate workflow runs when merging PRs to master.

Previously, merging a PR fired both 'push' and
'pull_request_target' simultaneously, causing two runs to
race for the gh-pages force-push. The loser failed with a
"cannot lock ref" error. With pull_request, only the push
event deploys on merge, while PRs still get build-tested.

Author: hakonhagland

.github/workflows/python_sphinx_docs.yml

@@ -4,9 +4,15 @@ on:
     push:
         branches:
           - '**'
-    pull_request_target:
-        types: closed
-        branches: master
+    # Use pull_request (not pull_request_target) to build-test docs on PRs targeting master.
+    # When a PR is merged, only the 'push' event triggers deployment to gh-pages.
+    # Using pull_request_target here previously caused a race condition: merging a PR fired
+    # both 'push' and 'pull_request_target' simultaneously, producing two concurrent runs
+    # that both tried to force-push to gh-pages. The loser would fail with:
+    #   "cannot lock ref 'refs/heads/gh-pages': is at <new-sha> but expected <old-sha>"
+    # See: https://github.com/OPM/opm-python-documentation/actions/runs/21938360885
+    pull_request:
+        branches: [master]
     repository_dispatch:
       types: [docstrings_common_updated, docstrings_simulators_updated]
 permissions: