* Correctly identify whitespace.
* Merge consecutive tokens where possible.
* Consistently use String.Double for the opening and closing quotation
  marks

Correctly identify whitespace.

Use a simpler expression to match comments in HTML/XML.

Part of pygments#1718.

wcag-contrast-ratio is now a build-time dependency for the docs.

Some dark styles did not define a color for every token type,
resulting in black text (the browser default for text) on dark
backgrounds (defined by the styles) unless the web page had some
CSS to remedy that like:

  body { color: white; background: black; }

We however don't want the readability of styles to rely on external CSS.

Part of pygments#1718. Fixes some unreadable styles reported in pygments#1526.

Part of pygments#1718.

fixes pygments#1962

…s#1961)

fixes pygments#1949

Alloy (http://alloytools.org/) facts can take a string token in addition
to an identifier. This is because the name isn't actually used for
facts, it's just for readability.

This is technically undocumented behavior, but it's widely used in
legacy specs and is handled this way in the official Alloy IDE.

Alloy 6 adds temporal logic to Alloy. This includes the following
keywords:

* Three new symbols: `'` (prime), `..` (step interval), and `;` (sequential composition)
* Five future-time keywords: `always`, `after`, `eventually`, `until`,
  `releases`
* Five past-time keywords: `historically`, `before`, `once`, `since`,
  `triggered`
* A new `steps` keyword for specifying a range to model check.

See here for full changes: https://alloytools.org/alloy6.html

Legacy alloy uses `'` to distinguish variables:

    all b, b': Foo | bar

But since Alloy 6 makes `'` into an operator, the new recommendation is
to use `"`:

    all b, b": Foo | bar

See
https://alloytools.org/alloy6.html#compatibility-with-pre-6-models.

The lexer currently reads `b"` as starting a string. This commit fixes
this by only having `"` begin a string if it's the start of a token.

Additionally, `'` is removed from identifiers.

Alloy 6 recommends changing `'` in legacy specs to `"`. Since this is a
breaking change, the golden test is updated to reflect that.

Alloy (http://alloytools.org/) facts can take a string token in addition
to an identifier. This is because the name isn't actually used for
facts, it's just for readability.

This is technically undocumented behavior, but it's widely used in
legacy specs and is handled this way in the official Alloy IDE.

Alloy 6 adds temporal logic to Alloy. This includes the following
keywords:

* Three new symbols: `'` (prime), `..` (step interval), and `;` (sequential composition)
* Five future-time keywords: `always`, `after`, `eventually`, `until`,
  `releases`
* Five past-time keywords: `historically`, `before`, `once`, `since`,
  `triggered`
* A new `steps` keyword for specifying a range to model check.

See here for full changes: https://alloytools.org/alloy6.html

Legacy alloy uses `'` to distinguish variables:

    all b, b': Foo | bar

But since Alloy 6 makes `'` into an operator, the new recommendation is
to use `"`:

    all b, b": Foo | bar

See
https://alloytools.org/alloy6.html#compatibility-with-pre-6-models.

The lexer currently reads `b"` as starting a string. This commit fixes
this by only having `"` begin a string if it's the start of a token.

Additionally, `'` is removed from identifiers.

Alloy 6 recommends changing `'` in legacy specs to `"`. Since this is a
breaking change, the golden test is updated to reflect that.

As mentioned in pygments#1963 (comment), the problem is that the rule matched . before .., so .. would never be matched. Fixed here.