libcore: Get rid of useless mem::forget wrapper.

[luqmana]

Fixes #15401.

[huonw]

I think these docs (and even the #[stable]) may need to go on the actual definition, not the reexport.

(@aturon would know for sure about the reexport.)

[brson]

Do intrinsics seamlessly cast to rust functions yet? This is a breaking change if not, and one I have argued against several times. How does rustdoc deal with this?

[huonw]

How does rustdoc deal with this?

Poorly, e.g. transmute is reexported in a similar manner: http://doc.rust-lang.org/master/std/mem/fn.transmute.html

[brson]

I see transmute has been done. I think we should consider it a bug if this exposing special rust-intrinsic semantics.

[alexcrichton]

Yes, I agree with @brson that this needs to be a [breaking-change], and with @huonw that the documentation/attribute should move to the actual declaration instead of the pub use (for now).

I do think, however, it's ok to land for now. RFC 5 will give intrinsics the same ABI as other rust functions, so it is going to be fixed on the horizon.

[brson]

I also agree it's ok to land as long as we're in agreement there are bugs.

[brson]

@bors seems to have not noticed the r+ here, but @alexcrichton did ask for a [breaking-change] tag.