Serialize workflow runs per PR (#47)

Two runs can act on the same PR concurrently: the push that finishes a
conflict resolution fires a `synchronize` event while the conflict label
is still attached (the label is the gate and only comes off at the end),
so the follow-up run races the one that triggered it. A per-PR
concurrency group queues the follow-up instead; `cancel-in-progress:
false` so a running update is never killed mid-mutation.

The group is keyed by PR number rather than repo-wide on purpose: GitHub
keeps only the newest *pending* run per group, so a repo-wide group
could silently drop the stack update for one merge when several land in
quick succession.

Also added to both README example workflows.

🤖 Generated with [Claude Code](https://claude.com/claude-code)

https://claude.ai/code/session_01JHvKryT4QUpHYdNq9YEQxX

---
_Generated by [Claude
Code](https://claude.ai/code/session_01JHvKryT4QUpHYdNq9YEQxX)_

---------

Co-authored-by: Claude <noreply@anthropic.com>

Author: Phlogistique

.github/workflows/update-pr-stack.yml

@@ -9,6 +9,13 @@ permissions:
   contents: write
   pull-requests: write
 
+# Serialize runs per PR: the push that finishes a conflict resolution fires a
+# synchronize event while the conflict label is still attached, so without a
+# group the follow-up run races the one that triggered it.
+concurrency:
+  group: update-pr-stack-${{ github.event.pull_request.number }}
+  cancel-in-progress: false
+
 jobs:
   update-pr-stack:
     runs-on: ubuntu-latest

README.md

@@ -80,6 +80,10 @@ on:
   pull_request:
     types: [closed, synchronize]
 
+concurrency:
+  group: update-pr-stack-${{ github.event.pull_request.number }}
+  cancel-in-progress: false
+
 jobs:
   update-pr-stack:
     runs-on: ubuntu-latest
@@ -111,6 +115,10 @@ permissions:
   contents: write
   pull-requests: write
 
+concurrency:
+  group: update-pr-stack-${{ github.event.pull_request.number }}
+  cancel-in-progress: false
+
 jobs:
   update-pr-stack:
     runs-on: ubuntu-latest

tests/test_e2e.sh

@@ -558,7 +558,7 @@ wait_for_synchronize_workflow() {
                 log_cmd gh run view "$target_run_id" --repo "$REPO_FULL_NAME" --log || echo >&2 "Could not fetch logs for run $target_run_id"
                 return 1
             fi
-        elif [[ "$run_status" == "queued" || "$run_status" == "in_progress" || "$run_status" == "waiting" ]]; then
+        elif [[ "$run_status" == "queued" || "$run_status" == "in_progress" || "$run_status" == "waiting" || "$run_status" == "pending" ]]; then
             echo >&2 "Workflow $target_run_id is $run_status. Sleeping $sleep_time seconds."
         else
             echo >&2 "Workflow $target_run_id has unexpected status: $run_status. Conclusion: $run_conclusion"
@@ -665,7 +665,7 @@ wait_for_workflow() {
                 log_cmd gh run view "$target_run_id" --repo "$REPO_FULL_NAME" --log || echo >&2 "Could not fetch logs for run $target_run_id"
                 return 1
             fi
-        elif [[ "$run_status" == "queued" || "$run_status" == "in_progress" || "$run_status" == "waiting" ]]; then
+        elif [[ "$run_status" == "queued" || "$run_status" == "in_progress" || "$run_status" == "waiting" || "$run_status" == "pending" ]]; then
             echo >&2 "Workflow $target_run_id is $run_status. Sleeping $sleep_time seconds."
         else
             echo >&2 "Workflow $target_run_id has unexpected status: $run_status. Conclusion: $run_conclusion"