Adds std::duration::Duration from/to Python conversions

[Tpt]

No description provided.

[Tpt]

We already support chrono::Duration, it might be nice to support std::duration::Duration for consistency.

Things to notice:

• std::duration::Duration only supports positive durations, an exception is raised when converting a negative timedelta
• a warning is created if precision is lost when converting from Duration to timedelta, opposite to chrono::Duration. Is it a good idea?

[adamreichold]

LGTM

[davidhewitt]

👍 likewise, thanks!

It's a little sad that Rust stdlib duration is strictly positive, but that's not really our design problem to worry about here.

[davidhewitt]

Hmm, there are some test failures?

• a warning is created if precision is lost when converting from Duration to timedelta, opposite to chrono::Duration. Is it a good idea?

Yeah I think that's fine to warn 👍

[Tpt]

Thank you for the review!

Hmm, there are some test failures?

My bad. It should be fixed now.

a warning is created if precision is lost when converting from Duration to timedelta, opposite to chrono::Duration. Is it a good idea?

Yeah I think that's fine to warn 👍

After thinking a bit more about it, I believe warning is a bad idea: it might overflow the warning logs if the conversion is done a lot of times. And it is likely there are cases where the user does not care about the precision loss (for example if the duration is an elapsed time between two timestamp). I would tend to just remove it (like chrono::Duration behaves) or introduce a static boolean to print it only once per process execution. I have removed the warning in this MR but I am happy to add it back if you feel it's useful.

[adamreichold]

@Tpt Argh, unused import on wasm32-wasi derailed the CI...

[adamreichold]

Pushed a fixup.