Better document sniffing and default format parameters

See https://github.com/wireservice/csvkit/issues/751#issuecomment-310811491 and https://github.com/wireservice/csvkit/issues/844

Describe how sniffing works for all tools. Also, add a troubleshooting section for errors like in #844 (closed).

Also, consider a global option to report what dialect was sniffed (interacts with agate, though).