Fix #19: lone surrogates should not be replaced by U+FFFD.

These can only occur in the input stream from arbitrary unicode
strings being passed into the parser (e.g., from script); no decoder
will emit them nowadays. This fixes us to match the current spec.

Author: gsnedders

tokenizer/unicodeCharsProblematic.test

@@ -2,22 +2,22 @@
 {"description": "Invalid Unicode character U+DFFF",
 "doubleEscaped":true,
 "input": "\\uDFFF",
-"output":["ParseError", ["Character", "\\uFFFD"]]},
+"output":["ParseError", ["Character", "\\uDFFF"]]},
 
 {"description": "Invalid Unicode character U+D800",
 "doubleEscaped":true,
 "input": "\\uD800",
-"output":["ParseError", ["Character", "\\uFFFD"]]},
+"output":["ParseError", ["Character", "\\uD800"]]},
 
 {"description": "Invalid Unicode character U+DFFF with valid preceding character",
 "doubleEscaped":true,
 "input": "a\\uDFFF",
-"output":[["Character", "a"], "ParseError", ["Character", "\\uFFFD"]]},
+"output":[["Character", "a"], "ParseError", ["Character", "\\uDFFF"]]},
 
 {"description": "Invalid Unicode character U+D800 with valid following character",
 "doubleEscaped":true,
 "input": "\\uD800a",
-"output":["ParseError", ["Character", "\\uFFFDa"]]},
+"output":["ParseError", ["Character", "\\uD800a"]]},
 
 {"description":"CR followed by U+0000",
 "input":"\r\u0000",