Replace macro impls in libstd/num/strconv.rs with generic impls

[Kimundi]

The impl macros in libstd/num/strconv.rs should get replaced by two generic impls for the not yet existing Integral and Floating traits once those traits exists, and once the compiler supports multiple generic impls against different traits.

[pnkfelix]

As the bug description says, this is blocked by revisions to the numeric traits, which I think are being tracked in Issue #4819.

[Kimundi]

Also, the bug might become irrelevant because I'm kinda working on re-replacing that central trait with macro-generated specialized versions for ints, uints and floats. But that hasn't landed yet, so for now this is still relevant.

[alexcrichton]

In theory the change to make is to write something like impl<T: Integer> NumStrConv for T, but this is currently blocked on some resolve bugs because if you do this for both T:Integer and T:Float you get multiple implementation of a trait errors.

[huonw]

Triage: the macros are still there; this is part of #6220.

[flaper87]

Triage: Updated the title and description s/libcore/libstd/ the macros are still there

[aturon]

@bjz do you still plan to work on this (and to hide strconv from the public API)?

[brendanzab]

Yes, I do, sorry for the delay.

[brson]

There are no numeric traits now and this won't happen.