zcsv

Copperbox revisions 3145 and 3146.

Initial work on a lexer that should be able to decode fields during parsing (rather than parsing to a string and using read [1]). The lexer is a "standard" two continuation parser monad, but it is quite stripped down (no source position tracking, no error messages) so hopefully it should be quite fast.

Currently the lexer isn't hooked up to the main CSV parser.

[1] Because strings are expected to be short, it's moot whether it will turn out to be faster than read. However, read is specialized to Haskell values so it still seems worthwhile to have my own implementation.