Prefer core::time::Duration over std::time::Duration

[mnmaita]

Objective

• While working on a bevy 0.15 project and then checking the current state of the main bevy branch to see why my autocomplete wasn't picking up the bevy::utils::Duration export I noticed that I couldn't find it anymore. As far as I can see Remove everything except Instant from bevy_utils::time #17158 removed this re-export, yet some parts of the code still use std::time::Duration. To standardize this I'm replacing all instances of std::time::Duration in favor of core::time::Duration. These are mostly example code and comments for what is worth.

Solution

• Replaced all instances of std::time::Duration in favor of core::time::Duration.

Testing

• I haven't tested this changeset thoroughly as I think its impact should be negligible. Might be good to perform an example run on CI though.

[alice-i-cecile]

Generally we don't worry about no_std in examples / test / benchmarks, which is why these are still around.

[bushrat011899]

Personally, I think there's no harm here and it is more consistent with how the Bevy crates are written now. However, I did propose this before and @mockersf raised a good point that this is not how the majority of Bevy users will write code, so it negatively affects the utility of the examples.

To be clear, I support this change, but this is at the very least controversial.

[mockersf]

Generally we don't worry about no_std in examples / test / benchmarks, which is why these are still around.

mostly seconding this, but I don't think it matters for benchmarks.

The change in example window/change_window_mode.rs is a pathological case of why this isn't really good:

            std::thread::sleep(core::time::Duration::from_secs(4));

We are using std threads there, why introduce confusion?