Leaderboard Deploy agents

When destroying objects asynchronously to avoid crashes from observer notifications or other synchronous destruction issues, ensure proper cleanup ordering and add defensive null checks to handle race conditions.

Key practices:

1. Move destruction to async context: Use task runners to defer destruction when synchronous cleanup causes crashes due to active observers or notifications
2. Reset pointers immediately: After moving objects to async destruction, immediately set the original pointer to null to prevent use-after-move
3. Add defensive null checks: In methods that might be called during async destruction, add null checks before accessing potentially destroyed objects
4. Explicit resource cleanup: For resources requiring manual shutdown (like dbus::Bus), ensure cleanup methods are called before destruction

Example from the discussions:

// Instead of synchronous destruction that crashes:
// managed_devtools_web_contents_.reset();

// Use async destruction:
embedder_message_dispatcher_.reset();
content::GetUIThreadTaskRunner({})->PostTask(
    FROM_HERE,
    base::BindOnce(
        [](std::unique_ptr<content::WebContents> web_contents) {},
        std::move(managed_devtools_web_contents_)));
managed_devtools_web_contents_ = nullptr;

// Add defensive checks in methods:
if (!GetDevToolsWebContents())
  return;

This pattern prevents crashes from premature destruction while maintaining system stability during async cleanup operations.