Thanks to @daurnimator for pointing this out on #1773
According to https://simonsapin.github.io/wtf-8/, Windows file paths are not necessarily valid utf-16le. When using Windows APIs the zig standard library should not panic or return an encoding error when encountering such file paths.

Zig's cross platform file system abstractions must handle this, plus unix file paths which are null terminated byte arrays.

Potentially we could decode into WTF-8 rather than UTF-8 on Windows, however then it may become subtle to know what kind of operations are allowed on file paths. For example, the algorithm for concatenation is different for each encoding.

Related: #534