Add "std::byte" support to stream

[gmlueck]

Since we encourage people to migrate from the old sycl::byte to
std::byte, I presume we want to support std::byte in the stream
class.

[ProGTX]

Interesting, are bytes streamable? The example in https://en.cppreference.com/w/cpp/types/byte suggests you need to convert it to an integer first.

[keryell]

Interesting, are bytes streamable? The example in https://en.cppreference.com/w/cpp/types/byte suggests you need to convert it to an integer first.

Good point.
This suggest not adding it to the stream interface.

[gmlueck]

Yes, I agree that this is a good point. I think there should still be some way for applications to emit the value of an std::byte to a sycl::stream, though. If we think of sycl::stream as an analog of <iostream>, then device code should do something like this:

sycl::stream sstream(/* ... */);

// ...

std::byte b{42};
sstream << std::to_integer<int>(b);

For that to be legal, we need to say someplace that std::to_integer<>() is safe to call in device code. We should also list the other non-member functions related to std::byte (https://en.cppreference.com/w/cpp/types/byte) as safe.

Should we add a new section that lists functions and types from the standard C++ library that are safe in device code? It seems like the typical implementation of the functions related to std::byte would not require any changes in order to be safe in device code, so listing these in the spec would likely not be an implementation burden. Over time, we will probably find other std functions that we want to add to this section.

[fraggamuffin]

stream op for byte does not align with C++
setup section initially for byteops( std C++ fns called from device code)

[keryell]

2021/09/02 SYCL Khronos committee meeting

Waiting for new wording from @gmlueck.
This PR is not ready for merging yet.

[fraggamuffin]

awaiting draft for NEXT or KHR