std::ascii: Provide a copyless [Ascii] -> str method.

[huonw]

This renames to_str_ascii to as_str_ascii and makes it non-copying,
which is possible now that strings no longer have a hidden extra
byte/null terminator.

Fixes #6120.

[Kimundi]

Replacing to_str_ascii with into_str is strange, because into_str is implemented inconsistent compared to to_str:

"foo".as_ascii().to_str() == "[f, o, o]" // because of generic impl of ToStr for &[T] with T: ToStr
"foo".as_ascii().into_str() == "foo" // Because into_str is implemented for &[Ascii]

I originally added the to_str_ascii method because of this issue: A &[Ascii] is both a ASCII string and a vector of ASCII codepoints, so there are two ways to convert it into a string.

The true fix here would be to have an str_ascii unsized type as a newtype wrapper around [Ascii], just like in theory str is a newtype wrapper around [u8] today.

.as_ascii() would then return a &str_ascii instead of a &[Ascii], and str_ascii could implement both ToStr and IntoStr as returning "foo" in the example above.

But because that's not possible yet, I'd keep to_str_ascii for now.

[huonw]

Replacing to_str_ascii with into_str is strange, because into_str is implemented inconsistent compared to to_str:

There is no mention of to_str in this PR. Replacing to_str_ascii with into_str is because it's converting (~[Ascii]) -> ~str; the old code using to_str_ascii would borrow that to &[Ascii] and allocate itself a new ~str. (into_str and to_str_ascii have the same format, i.e. "foo"... I don't see why being inconsistent with to_str is directly relevant?)

But because that's not possible yet, I'd keep to_str_ascii for now.

This is actually keeping it, just converting it to be &[Ascii] -> &str rather than &[Ascii] -> ~str (i.e. not doing a malloc internally) and changing its name to match the convention.

(I agree that having a proper ascii string is the "correct" fix.)

[Kimundi]

Sorry about the confusion, I misunderstood a few things. :)

What I wanted to say is this: Giving &[Ascii] a as_str_ascii method is perfectly fine, but seeing you replacing the old to_str_ascii calls with into_str calls, I reached the conclusion that ToStrConsume should not be implemented on ~[Ascii] at all because it leads to an inconsistency with the output of to_str(). Instead, there should be an into_str_ascii method.

However, seeing how the proper fix of having an ascii string type would allow to consistently implement both ToStr and ToStrConsume, I now think that keeping the current situation is fine too.

(Also, I think I introduced the ToStrConsume trait for use on a ~[Ascii] to begin with, and just forgot, so you really have all right to blame me for confusing you ;))

[Kimundi]

Ah, one more thing: While you're changing stuff in the ascii module anyway, could you remove the doc comment wording that talks about a null terminator? As neither str nor [Ascii] use such a thing now, having the docs talk about them is just confusing and wrong.