core: impl LowerHex and UpperHex for [u8]

[richo]

This is a fairly speculative PR, since I have a bunch of open questions:

• Is it actually as much work as it looks to implement this without the Debug bound? (Is that desirable? Happy to do it if it would get merged)
• Is marking this stable OK? It seems pretty minor
• More subjectively, this seemed to be right on the cusp of worth macro-ifying. I suspect that it's actually worth doing this for all of [u8], [u16] and friends, so I presume that I should actually macro it?

[rust-highfive]

r? @aturon

(rust_highfive has picked a reviewer for you, use r? to override)

[richo]

@aturon: Feel free to ignore the details of the implementation, I ran the tests overnight and it's broken, but I'd love feedback on the concept/the questions in the PR.

[alexcrichton]

This was originally proposed in #19670 (in a more general form) which was PR'd in #25319, but was closed due to concerns about the usefulness.

I feel like if we take this route we'd want to take the more general approach, but I think I still feel that this may not want to be in the standard library just yet.

[richo]

Interesting, thanks for the background. Based on your comment on #25319, would I be better off championing this in an RFC or an internals thread? I totally understand the conservative approach, but I'm pretty convinced that this, or something that looks a lot like this, is going to be something that crops up again and again.

As an example, I work with binary data a lot, and having no easy way to spit out internal representation in a way that can be easily compared to xxd introduces a lot of copying and pasting into a python repl into my workflow that I'd really rather avoid.

[alexcrichton]

would I be better off championing this in an RFC or an internals thread?

Yeah I think that'd be a good place to start a conversation for this, thanks!

[vi]

Are there any updates for the feature?

It's not convenient to copy and paste things from [191, 0, 24, 33, into calculator (or to complexify temporary println! with a loop, or wait for crates.io index update because of silly mini-dependency like hex_slice) just to see if some bit is set in n'th byte.

[stevenroose]

I always thought this was impossible because it would conflict with impl<T: LowerHex> LowerHex for [T], but that implementation doesn't exist. Why on earth can't we have LowerHex for [u8] and Vec<u8> then? Basically AsRef<[u8]>??