This is a conservative implementation for issue #36. I don't always agree with things stripping the UTF BOM on a file, but it seems it may be useful for the CSV parser. This will strip the BOM from a stream only when the stream's encoding is set to UTF-8 and the BOM is the first character in the first data event (which should be the beginning of the file 99% of the time).
Skip UTF BOM from first data event on UTF-8 decoded stream
This is how I was planning to do it. Look perfect to me.