Bug: UTF-8 characters across chunk boundaries get corrupted

[yetzt]

When the bytes of a UTF-8 character happen to get split across chunks when requesting a resource that is treated as text and not parsed, StreamDecoder._transform will corrupt it by applying iconv.decode on the chunk and thereby replacing the partial unicode character with a replacement character.

needle v3.0.0, reproduce:

require("needle").get("https://data.interaktiv.cloud.funkedigital.de/wahl/example/nw/erg_05994.xml", { parse: false }, function(err, resp, data){
	console.log(data.split('\n')[4379]); // line 4380
});

Suggested solutions:

• Use iconv.decodeStream instead of iconv.decode
• Complete chunk collection first, then use iconv.decode on complete body
• Check if the last byte of a chunk is the first byte of a UTF-8-Character, and if so remove it from the current chunk and prepend it to the next chunk.