Fix distributed bookmark queue resume race

[Copilot]

Purpose

Prevent parent workflows from remaining suspended after completed child workflows in distributed runtimes.

---

Scope

Select one primary concern:

• Bug fix (behavior change)
• Refactor (no behavior change)
• Documentation update
• Formatting / code cleanup
• Dependency / build update
• New feature

If this PR includes multiple unrelated concerns, please split it before requesting review.

---

Description

Problem

DispatchWorkflow(WaitForCompletion = true) can lose the bookmark queue wake-up in multi-node deployments when the signaled node cannot acquire the distributed queue lock. With the previous 1-minute default purge TTL, the queued resume item could be deleted before the next retry sweep.

Solution

• Retry on lock miss

  • DistributedBookmarkQueueWorker now re-signals after a short delay when the distributed lock is unavailable.
  • This keeps queued bookmark items eligible for processing instead of relying only on the next periodic trigger.

• Safer purge window

  • Default bookmark queue purge TTL increased from 1 minute to 1 hour.

• Regression coverage

  • Added unit coverage for lock-miss re-signaling.
  • Added unit coverage for the default purge TTL.

if (handle == null)
{
    Logger.LogDebug("Could not acquire lock for distributed bookmark queue worker...");
    await Task.Delay(RetryDelay, cancellationToken);
    await Signaler.TriggerAsync(cancellationToken);
    return;
}

---

Verification

Reviewers can verify the behavior through the added unit tests and by inspecting the distributed worker lock-miss path.

Steps:

1. Run Elsa.Workflows.Runtime.UnitTests.
2. Review DistributedBookmarkQueueWorkerTests.
3. Review BookmarkQueuePurgeOptionsTests.

Expected outcome:

Queued bookmark resume items are retried after distributed lock contention and are not purged before normal retry windows.

---

Screenshots / Recordings (if applicable)

N/A

---

Commit Convention

We recommend using conventional commit prefixes:

• fix: – Bug fixes (behavior change)
• feat: – New features
• refactor: – Code changes without behavior change
• docs: – Documentation updates
• chore: – Maintenance, tooling, or dependency updates
• test: – Test additions or modifications

Clear commit messages make reviews easier and history more meaningful.

---

Checklist

• The PR is focused on a single concern
• Commit messages follow the recommended convention
• Tests added or updated (if applicable)
• Documentation updated (if applicable)
• No unrelated cleanup included
• All tests pass