LibCore: Move event queueing to a per-thread event queue

Instead of juggling events between individual instances of
Core::EventLoop, move queueing and processing to a separate per-thread
queue (ThreadEventQueue).

Userland/Libraries/LibCore/ThreadEventQueue.h

@@ -0,0 +1,44 @@
+/*
+ * Copyright (c) 2023, Andreas Kling <kling@serenityos.org>
+ *
+ * SPDX-License-Identifier: BSD-2-Clause
+ */
+
+#pragma once
+
+#include <AK/NonnullOwnPtr.h>
+#include <AK/OwnPtr.h>
+
+namespace Core {
+
+// Per-thread global event queue. This is where events are queued for the EventLoop to process.
+// There is only one ThreadEventQueue per thread, and it is accessed via ThreadEventQueue::current().
+// It is allowed to post events to other threads' event queues.
+class ThreadEventQueue {
+    AK_MAKE_NONCOPYABLE(ThreadEventQueue);
+    AK_MAKE_NONMOVABLE(ThreadEventQueue);
+
+public:
+    static ThreadEventQueue& current();
+
+    // Process all queued events. Returns the number of events that were processed.
+    size_t process();
+
+    // Posts an event to the event queue.
+    void post_event(Object& receiver, NonnullOwnPtr<Event>);
+
+    // Used by Threading::BackgroundAction.
+    void add_job(NonnullRefPtr<Promise<NonnullRefPtr<Object>>>);
+
+    // Returns true if there are events waiting to be flushed.
+    bool has_pending_events() const;
+
+private:
+    ThreadEventQueue();
+    ~ThreadEventQueue();
+
+    struct Private;
+    OwnPtr<Private> m_private;
+};
+
+}