Fix tracing spans for system

[stepancheg]

Objective

Spans are not recorded correctly for system,

Solution

Do not create spans prematurely.

Test

Modify logs example:

diff --git a/examples/app/logs.rs b/examples/app/logs.rs
index 797fd85fc..787f4d381 100644
--- a/examples/app/logs.rs
+++ b/examples/app/logs.rs
@@ -25,6 +25,8 @@ fn log_system() {
     warn!("something bad happened that isn't a failure, but thats worth calling out");
     error!("something failed");

+    panic!("test");
+
     // by default, trace and debug logs are ignored because they are "noisy"
     // you can control what level is logged by setting up the LogPlugin
     // alternatively you can set the log level via the RUST_LOG=LEVEL environment variable

Apply another fix from separate PR #11170.

Run command:

cargo run --example logs --features=trace

Before:

2024-01-01T14:36:02.900476Z  INFO system{name="logs::log_once_system"}: logs: doing expensive things
2024-01-01T14:36:02.900449Z ERROR system{name="logs::log_system"}: logs: something failed
   0: bevy_ecs::system::function_system::system
           with name="logs::log_system"
             at crates/bevy_ecs/src/system/function_system.rs:47

After:

2024-01-01T14:34:32.119761Z ERROR bevy_app:winit event_handler:update:main app:schedule{name=Main}:system{name="bevy_app::main_schedule::Main::run_main"}:schedule{name=Update}:system_task{name="logs::log_system"}:system{name="logs::log_system"}: logs: something failed
2024-01-01T14:34:32.119823Z  INFO bevy_app:winit event_handler:update:main app:schedule{name=Main}:system{name="bevy_app::main_schedule::Main::run_main"}:schedule{name=Update}:system_task{name="logs::log_once_system"}:system{name="logs::log_once_system"}: logs: doing expensive things
   0: bevy_ecs::system::function_system::system
           with name="logs::log_system"
             at crates/bevy_ecs/src/system/function_system.rs:479
   1: bevy_ecs::schedule::executor::multi_threaded::system_task
           with name="logs::log_system"
             at crates/bevy_ecs/src/schedule/executor/multi_threaded.rs:156
   2: bevy_ecs::schedule::schedule::schedule
           with name=Update
             at crates/bevy_ecs/src/schedule/schedule.rs:291
   3: bevy_ecs::system::exclusive_function_system::system
           with name="bevy_app::main_schedule::Main::run_main"
             at crates/bevy_ecs/src/system/exclusive_function_system.rs:110
   4: bevy_ecs::schedule::schedule::schedule
           with name=Main
             at crates/bevy_ecs/src/schedule/schedule.rs:291
   5: bevy_app::app::main app
             at crates/bevy_app/src/app.rs:261
   6: bevy_app::app::update
             at crates/bevy_app/src/app.rs:258
   7: bevy_winit::winit event_handler
             at crates/bevy_winit/src/lib.rs:345
   8: bevy_app::app::bevy_app
             at crates/bevy_app/src/app.rs:299

Changelog

• Fix tracing spans for system.

[james7132]

This effectively reverts #9390 and re-opens #4892. This also recollects effectively the same stack trace for every system, so it's generally more noise than it's actually beneficial to users.

With the effort @hymm has spent ensuring we get accurate traces while profiling Bevy apps, I am very much against this change as is.

[alice-i-cecile]

Closing this out then.