In Python 3, RawTokenFormatter would output non-ASCII for non-ASCII
input, and RawTokenLexer would throw Unicode-related exceptions for
ASCII or non-ASCII input; fix them.  Also, handle all exceptions, so
that callers who find RawTokenLexer via get_lexer_by_name on user
input don’t unexpectedly get a lexer that throws exceptions.

Signed-off-by: Anders Kaseorg <andersk@mit.edu>

* Add (all?) Matlab built-ins

See also https://de.mathworks.com/help/matlab/referencelist.html

* Adjust pytest goldens

* Add CDDL lexer (thanks to Fabian Neumann)

* Add CDDL to mappings

* Fix inline flag in CDDL regex

* Update AUTHORS

* Fix explosive backtracking

* Comment invalid CDDL syntax for automated tests

* Update following Georg Brandl's review

* Update tests for CDDL to new framework

* Pylint pass

* Update links to CDDL RFC

* Update copyright header

* Solve regexlint issues in CDDL parser

* Add link to CDDL in documentation

* Fix #1698.

* Change default line number styling.

Instead of hardcoding black/light gray, use transparent/inherit which
makes it "neutral" out of the box. linenospecial is still hard-coded and
may look wrong, but that needs fixing in the style.

* Prettify the test files.

* Fix incorrect wrapping of TD linenos.

The padding would get applied twice to "special" lines, once via
<pre>, once via the <span>, which was not consistent with inline
line numbers where the .special style would override the default
style.

To fix this, we now emit "normal" and "special" lines for td.linenos
line numbers, and the normal style is applied to "normal". This brings
td.linenos closer to inline line numbers, and also makes the styling
between noclasses=True and noclasses=False consistent.

* Document the line number styling changes.