Optimize NativeString: add UTF-8 conversion fast path

[ValoChet]

JavaScript strings are stored either as Latin-1 or UTF-16, and uWS always convert them to UTF-8.

For small strings, we can avoid the extra Utf8Length() pass by using a safe worst-case allocation size:

• Latin-1 -> UTF-8 = 2 * size in worst case
• UTF-16 -> UTF-8 = 3 * size in worst case

If this worst-case size fits in the pool buffer, we can call WriteUtf8() directly.
It writes the UTF-8 data into the pool and returns the exact number of bytes written.
The pool offset is then advanced by the actual number of bytes used.

This should cover most strings thanks to the generous 128 KB pool.
I the string is too large, it fallback calculating the exact size.

Test case: 10 x writeHeader(32 bytes, 32 bytes)

Input type	Before	After	Change
Binary	152k req/sec	152k req/sec	+0%
Latin-1	147k req/sec	151k req/sec	+3%
UTF-16	127k req/sec	146k req/sec	+15%