When multiple non US-ASCII chars are split over several chunks, Utf8ByteBufCharsetDecoder crashes with BufferOverflowException. This happens because we don't clear the `splitCharBuffer` in-between and only do so once decoding is done.