Refactor system diagnostics to use a single task

[dloukadakis]

Objective

Fixes #20802

Solution

System information diagnostics now uses a single task that wakes up every
time the First schedule runs. The task checks if enough time has
passed since the last refresh. If enough time has passed, it refreshes
the system information and sends it to a channel. The
read_diagonstic_task system then reads the system information from the
channel to add diagnostic data.

Testing

• Did you test these changes? If so, how?
  I used the log diagnostics example and compare before the changes and after the changes to ensure it works similarly or better than before.
• How can other people (reviewers) test your changes? Is there anything specific they need to know?

cargo run --example log_diagnostics

• If relevant, what platforms did you test these changes on, and are there any important ones you can't test?
  Linux

[alice-i-cecile]

This solves the same problem as #20845: we should pick one and coalesce our efforts :)

[NthTensor]

I'll have to compare this with the other PR. It would be good for you too read over the comments on that one.

[dloukadakis]

I just realized, I will review #20845 as well.

[dloukadakis]

I measured the CPU usage between main, #20852 and #20845 letting it run for 20 seconds per test.

#20852 (this pull request)

2025-09-03T18:56:21.619919Z  INFO bevy_diagnostic: system/cpu_usage :    9.591195%  (avg 9.572284%)
2025-09-03T18:56:21.619952Z  INFO bevy_diagnostic: process/cpu_usage:    9.119496%  (avg 8.990516%)

#20845

It's not reporting correctly, my system CPU usage is 22%.

2025-09-03T19:01:52.360746Z  INFO bevy_diagnostic: system/cpu_usage :    1.999126%  (avg 1.998040%)
2025-09-03T19:01:52.360777Z  INFO bevy_diagnostic: process/cpu_usage:    0.000000%  (avg 0.000000%)

It works, when we fix https://github.com/bevyengine/bevy/pull/20845/files#r2319871486 , my system CPU usage drops to 16%, and it starts reporting correctly.

2025-09-03T19:05:57.686947Z  INFO bevy_diagnostic: system/cpu_usage :   16.323296%  (avg 16.180935%)
2025-09-03T19:05:57.686989Z  INFO bevy_diagnostic: process/cpu_usage:   16.006338%  (avg 15.766926%)

main (0ffd9d6)

2025-09-03T18:53:54.654583Z  INFO bevy_diagnostic: system/cpu_usage :   10.244648%  (avg 9.765763%)
2025-09-03T18:53:54.654623Z  INFO bevy_diagnostic: process/cpu_usage:    9.021406%  (avg 9.076174%)

[Olle-Lukowski]

Let's go with this solution, I wrote my PR quite quickly honestly. And clearly that did not help with the quality. This looks better, and it's clear @dloukadakis has thoroughly tested this.

[NthTensor]

Looks great. This pattern, keeping a task around which we wake every frame, is very interesting. I wonder if there are other places within the engine where this would be useful.

Since this uses some bits of async that people may not be familiar with, I'd really appreciate a small doc comment that explains what's going on. Maybe on the plugin itself?

[NthTensor]

Fantastic work here, thank you.