Fix effects triggered from async tasks; improve work scheduling docs (#2370)

* outline work scheduling

* move scheduling code from dirty_scope to scheduler

* introduce queue_effect

* Run effects after all tasks are polled if there are no components rerendered

* add a new test for async effects

packages/hooks/src/use_effect.rs

@@ -41,25 +41,29 @@ use crate::use_callback;
 /// ```
 #[track_caller]
 pub fn use_effect(callback: impl FnMut() + 'static) -> Effect {
-    // let mut run_effect = use_hook(|| CopyValue::new(true));
-    // use_hook_did_run(move |did_run| run_effect.set(did_run));
-
     let callback = use_callback(callback);
 
     let location = std::panic::Location::caller();
 
     use_hook(|| {
+        // Inside the effect, we track any reads so that we can rerun the effect if a value the effect reads changes
         let (rc, mut changed) = ReactiveContext::new_with_origin(location);
+        // Spawn a task that will run the effect when:
+        // 1) The component is first run
+        // 2) The effect is rerun due to an async read at any time
+        // 3) The effect is rerun in the same tick that the component is rerun: we need to wait for the component to rerun before we can run the effect again
+        let queue_effect_for_next_render = move || {
+            queue_effect(move || rc.run_in(&*callback));
+        };
+
+        queue_effect_for_next_render();
         spawn(async move {
             loop {
-                // Run the effect
-                rc.run_in(&*callback);
-
                 // Wait for context to change
                 let _ = changed.next().await;
 
-                // Wait for the dom the be finished with sync work
-                wait_for_next_render().await;
+                // Run the effect
+                queue_effect_for_next_render();
             }
         });
         Effect { rc }