Avoid allocations on os_str_{from,to}_bytes on Windows

[glandium]

When dealing with valid utf-8, which is expected to be the most common case, we can just use the data as is. The downside is having to go over the string twice if it's not valid utf-8.

[dylni]

Thanks for the PR! I've considered this, but I don't think it's the right approach. This would add an unavoidable performance penalty to every conversion with non-UTF-8 bytes, rather than letting the user decide if UTF-8 should be parsed first.

However, it should be possible to do better. For os_str_from_bytes, this crate can easily tell if the bytes are valid UTF-8 while encoding, so I have an idea to avoid some work. For os_str_to_bytes, I would prefer to wait until this ACP is accepted to avoid the penalty.

Therefore, I am going to close this PR, as I do not plan to take this approach. Let me know if you have any additional input, and thank you for the suggestion!