std.meta: add comptimeKnown function

[SeanTheGleaming]

This function returns whether a value is comptime known.

I felt this function necessary to add to std.meta because it could be used to solve a few issues in the standard library and is overall just very useful for users as well. Off the top of my head, I can think of two immediate use cases in the standard library. Firstly, std.zig.c_builtins.__builtin_constant_p could be updated to work like it does in C rather than always returning 0, and more importantly, in #21498 there is some discussion around std.StaticStringMap regarding whether we should trade off the ability to have runtime keys for some strong optimizations which can be performed with comptime-known maps. With this function, we can have it both ways, and automatically choose whether to use the current implementation which can work for runtime maps, or the proposed optimized version which only works on comptime-known maps. Some follow up work may be considered to implement these changes, but for now I am just keeping it simple and just posting the function itself.

[alexrp]

The std.zig.c_builtins.__builtin_constant_p use case seems compelling to me at least. I suspect @andrewrk will have thoughts on this though, given the debate around @inComptime.

[alexrp]

Hmm, that said, I seem to vaguely recall @mlugg wanting to change the semantics around implicit comptimeness of tuple fields. Is that right? If so, we may not want to commit to this API unless there's an alternative implementation strategy.

[mlugg]

Hmm, that said, I seem to vaguely recall @mlugg wanting to change the semantics around implicit comptimeness of tuple fields. Is that right? If so, we may not want to commit to this API unless there's an alternative implementation strategy.

What I really want to do is remove comptime fields, but I'm not sure whether that's viable. I think this will continue working as long as comptime fields exist.

However... it's a blatant hack. I think if there's sufficient justification to warrant this functionality (somewhat debatable at this point in time IMO), we should just have a builtin for it.