Characters with more than one byte of data cause a bad request

[corymharper]

Take the following:

'☺'.length; // -> 1

new TextEncoder().encode('☺').byteLength; // -> 3

The smiley character has a length of one, but uses three bytes of memory. In the current implementation we calculate the Content-Length header using the length of the string, and for most characters that works, but there are some special characters that use multiple bytes of memory, this makes the header not match the actual amount of information passed, which causes a bad request.