no_std support on nightly

[FlorianUekermann]

As mentioned by #196, nightly now has the error_in_core feature. This make no_std support pretty straightforward (just replace std with core in a few places).

Some tests don't work on on no_std to keep the commit small for now. Maybe this is fine until error_in_core is merged, because there are only 5 lines in #[cfg(not(feature = "std"))] guards, so running the tests with std disabled hardly increases coverage anyway

Once error_in_core is in stable, even this little trick:

#[doc(hidden)]
pub mod __private {
    #[cfg(not(feature = "std"))]
    pub use core::error;
    #[cfg(feature = "std")]
    pub use std::error;

can be removed, or not, depending on how important the MSRV guarantees are.

Thoughts?

[FlorianUekermann]

I just rebased. @dtolnay do you have any comments on this?

[OverOrion]

This PR has been open for quite a while, are you going to review it @dtolnay anytime soon?

[dzmitry-lahoda]

@dtolnay what would be prerequisute for merge please ? thank you a lot

[dtolnay]

I would prefer to consider this only after it works on stable rustc.

Thanks anyway for the PR!

[dzmitry-lahoda]

okey, will continue use fork. btw this pr does not required nightly on std? so other users intact, only who want no std will be nightly?

[jordens]

I found this after going down the same road.

Here's a version that has all tests passing on both std and no_std, (compiletests can also be made to run on no_std but require a verbose splattering of cfg(error_in_core) plus overwrite of the output):

master...quartiq:thiserror:no-std