Tweak rust::Slice to become a C++20 contiguous_range.

[anforowicz]

The main missing piece was that
https://en.cppreference.com/w/cpp/iterator/random_access_iterator requires that "(a + n) is equal to (n + a)", but Slice::iterator only supported a + n and didn't support n + a.

[anforowicz]

PTAL?

Notes:

• Motivation: I learned recently that before this commit base::span<T> couldn’t be easily constructed out of rust::Slice<T>, because base::span’s constructor in Chromium requires the input to be a contiguous_range (see here and here).
• I wasn’t sure where to add static_asserts (in cxx.h vs in tests/ffi/[tests.cc](http://tests.cc/) vs somewhere else). I think having them in cxx.h is useful because it also serves as documentation. (Saying that Slice is a contiguous_range felt like a relatively low-level detail so I didn’t add that to book/src/binding/slice.md.)
• IIUC cargo test builds things in C++11 mode (?), so as-is there are no regression tests within cxx… :-/. But hopefully the new static_asserts will make it easy to detect and diagnose regressions if they get imported into C++20-based clients (like Chromium). So maybe this is good enough…
  • Ideally CI would test things with C++11, C++17, C++20, etc. I may try to do this as a follow-up, but I feel that I don’t know what I am doing, so let me try that in a separate PR. I tried making some progress here: https://github.com/anforowicz/cxx/tree/cxx20-coverage

[anforowicz]

/cc @zetafunction (many thanks for your help with the C++ templates :-) )

[dtolnay]

Thank you!

Yes I would appreciate a follow-up PR with CI. What you linked looks good so far. I'd be happy to have a distinct job for every C++ standard since C++11, even if there's no current intentional difference between some of them in this library.