The HTTP lexer only supports requests where the HTTP version is either
1.0 or 1.1. This means that HTTP/0.9 and HTTP/2 cause errors to be
generated. This commit replaces the version checks with general numeric checks.

ES6 added two regex flags to JavaScript's regular expressions: `u` and
`y`. This commit adds support for lexing these flags correctly.

This commit adds a lexer for Haxe.

This commit adds a lexer for Hive Query Language (a dialect of SQL). 

In addition, it updates the SQL lexer to move certain keywords to a
types category and tokenises these keywords with the `Name::Builtin`
token (this is consistent with the approach of Pygments and Chroma).

For keywords that can be used both in a type specification and as a
function (e.g. `COLLATE`), the categorisation was based on the most
common use of the keyword.

Terraform 0.12 adds support for first-class expressions. Prior versions
of Terraform required all non-literal expressions to be included as
interpolated sequences inside strings.

In addition, the demo and visual samples have been rationalised. This
will assist in future updates to the Terraform lexer.

To avoid instantiating a large number of strings, Rouge currently saves
the list of built-in keywords for MatLab to a file that is only read
when the MatLab lexer is in use. The keywords that are in this file are
generated by a Rake task that extracts the keywords from the official
MatLab documentation.

Since the time that the Rake task was created, MathWorks (the creators
of MatLab) have changed the way this page works such that it requires
JavaScript to load. This breaks the Rake task.

As an alternative, this commit includes the list of built-in keywords in
a YAML file. This is consistent with some of the other lexers that have
a large number of keywords (e.g. Apache lexer). It will, unfortunately,
require copying and pasting for future updates.

If there is more than one space before a single-line comment, the Q
lexer does not tokenise the comment correctly. This commit fixes that
bug.

…lexer (#1308)

The Elixir lexer does not properly handle escaping and interpolating in
string (`"foo"`) and charlist (`'foo'`) literals.

This PR adds general escaping rules for use in strings as well as adding
support for escaping and interpolating to charlist literals. It also
rationalises the states used for string literals. Examples to confirm
correct behaviour have been added to the visual sample.

This commit adds a lexer for embedded Elixir.