feat: Add memory.dataUTF8() built-in function

[geraintluff]

Implements #2972, providing a way to embed UTF-8 string constants.

Rationale

UTF-16 strings are naturally embedded in the data section if you use a string constant ("...") in your code.

If what you need is a UTF-8 string, then storing as WTF-16 wastes binary space and execution time.

Current workaround

You can currently achieve the same result by using memory.data<u8>([..., 0]) with the appropriate UTF-8 bytes, but reading/changing/reviewing this code is more difficult.

Changes

• Declare memory.dataUTF8(<string constant>) function in standard library
• Implement the built-in in the compiler
• Add test to tests/compiler/memory.ts

npm run check shows no errors for the files I touched. (Existing errors in tests/import/index.ts are unrelated.)

• I've read the contributing guidelines
• I've added my name and email to the NOTICE file

[geraintluff]

This FIXME is copied from the memory.data<T> implementation just above. Tbh I don't understand it, but if it's a problem there, then it's a problem here.

[geraintluff]

If you want to see why I proposed this: here's a diff

[geraintluff]

@dcodeIO I wrote a utf16ToUtf8() helper, checked portability using npm run bootstrap. It's a single pass, so allocates the most pessimistic buffer size (1.5x the WTF16 length in bytes), and then trims to size afterwards.

It silently encodes unpaired surrogates (WTF16) as code-points, which is the same behaviour as String.UTF8.encodeUnsafe() under the default ErrorMode.WTF8.