support multiline CoffeeScript comments

[davidchambers]

This pull request adds support for multiline CoffeeScript comments.

CoffeeScript comments are now extracted by using CoffeeScript.helpers.extractAllCommentTokens in conjunction with CoffeeScript.tokens (previously we used a regular expression, /^([ \t]*)(.*)\n?/gm). Although comment tokens have a here property, I decided not to use it when I realized that CoffeeScript supports inline ### comments. Instead, we use locationData.last_line > locationData.first_line to determine whether a comment spans multiple lines.

rewriteCoffee is now very similar to rewriteJs, at a high level:

$ awk '$2 ~ /rewrite(Js|Coffee)/ { print $2; look = 1 } $0 == "};" { look = 0 } look && $1 == "//" && $2 ~ /:$/ { print }' lib/doctest.js
rewriteJs
  // 1: Parse source text to extract comments
  // 2: Preserve source text between comments
  // 3: Extract prefixed comment lines
  // 4: Coalesce related input and output lines
  // 5: Convert doctests to source text
  // 6: Sort verbatim and generated source text by original offsets
  // 7: Concatenate source text
rewriteCoffee
  // 1: Lex source text to extract comments
  // 2: Preserve source text between comments
  // 3: Extract prefixed comment lines
  // 4: Coalesce related input and output lines
  // 5: Convert doctests to source text
  // 6: Sort verbatim and generated source text by original offsets
  // 7: Concatenate source text

The only significant difference is that rewriteJs parses the source text whereas rewriteCoffee only lexes it.

rewriteJs now extracts lines from multiline comments with String.prototype.indexOf and String.prototype.slice rather than String.prototype.split.

[davidchambers]

This should have been offset += text.length + '\n'.length; (it was not actually a problem since offsets are only used for ordering).