Feature suggestion: allow tools like hashcsv to handle mangled rows

For some kinds of bad rows (eg. incorrect numbers of fields), it may be useful to allow hashing. For example, I was hoping to use it to work around #8, but I can't hash an unscrubbed csv... 