format: Fix unicode escapes in string literals

[avh4]

Fixes #191

This is a little bit tricky because the AST contains the string values in a format that is optimized for outputting the compiled javascript (it is a fully-assembled ByteArray# in UTF8 encoding, with javascript escape sequences, also with unicode escapes normalized to use only 4-digit hex codes). Perhaps a more ideal option would be to refactor the parser and AST to retain the String.Chunks that are currently used internally in Gren.String, but that seemed complicated to do. So instead this PR just has the formatter try to re-parse the string to recover the information needed to format it properly.

(elm-format hasn't had to deal with this yet, since it is still based on an older version of Elm's parser, so the AST there contains the actual string that is represented, instead of the javascript-escaped version.)

[robinheghan]

Nice work!

[robinheghan]

It's a little unfortunate that the parser is optimized for the codegen output, but fixing this in the formatter seems to be the most low-risk way of fixing this.