The new lexer matches the TOML spec much more closely.

User-visible differences should be these:

* Add MIME type
* Highlight string escapes
* Recognize \uXXXX and \UXXXX escapes
* Also recognize booleans if they are followed by a comment
* Fix single quotes inside multiline literal strings (closes pygments#2488)
* Prevent multiline literal strings from eating comments
* Add multiline basic strings (""")
* Improve datetime recognition: recognize times without
  dates, dates without times and datetimes without time zone;
  allow sub-millisecond precision
* Recognize floats with exponents (they used not to be recognized
  when having a decimal point)
* Recognize binary, octal and hex literals
* Recognize strings inside table headers
* Recognize table headers followed by comments
* Don't parse sequences of digits as integers when they
  are actually keys

Includes several new tests, most of which were not working before.