add tracing spans for parallel executor and system overhead (#3416)

This PR adds tracing spans for the parallel executor and system overhead.

![image](https://user-images.githubusercontent.com/2180432/147172747-b78026e3-1c30-4120-92c8-693c6f1564cd.png)

Author: hymm

crates/bevy_ecs/src/schedule/executor_parallel.rs

@@ -6,6 +6,8 @@ use crate::{
 };
 use async_channel::{Receiver, Sender};
 use bevy_tasks::{ComputeTaskPool, Scope, TaskPool};
+#[cfg(feature = "trace")]
+use bevy_utils::tracing::Instrument;
 use fixedbitset::FixedBitSet;
 
 #[cfg(test)]
@@ -119,7 +121,7 @@ impl ParallelSystemExecutor for ParallelExecutor {
             .clone();
         compute_pool.scope(|scope| {
             self.prepare_systems(scope, systems, world);
-            scope.spawn(async {
+            let parallel_executor = async {
                 // All systems have been ran if there are no queued or running systems.
                 while 0 != self.queued.count_ones(..) + self.running.count_ones(..) {
                     self.process_queued_systems().await;
@@ -141,7 +143,12 @@ impl ParallelSystemExecutor for ParallelExecutor {
                     }
                     self.update_counters_and_queue_systems();
                 }
-            });
+            };
+            #[cfg(feature = "trace")]
+            let span = bevy_utils::tracing::info_span!("parallel executor");
+            #[cfg(feature = "trace")]
+            let parallel_executor = parallel_executor.instrument(span);
+            scope.spawn(parallel_executor);
         });
     }
 }
@@ -194,6 +201,9 @@ impl ParallelExecutor {
                 let system = system.system_mut();
                 #[cfg(feature = "trace")] // NB: outside the task to get the TLS current span
                 let system_span = bevy_utils::tracing::info_span!("system", name = &*system.name());
+                #[cfg(feature = "trace")]
+                let overhead_span =
+                    bevy_utils::tracing::info_span!("system overhead", name = &*system.name());
                 let task = async move {
                     start_receiver
                         .recv()
@@ -209,6 +219,9 @@ impl ParallelExecutor {
                         .await
                         .unwrap_or_else(|error| unreachable!(error));
                 };
+
+                #[cfg(feature = "trace")]
+                let task = task.instrument(overhead_span);
                 if system_data.is_send {
                     scope.spawn(task);
                 } else {