This fixes an issue where in code like this:

```
int foo(float bar) // hello() {}
```

The lexer would match `(float bar) // hello()`
as the parameters of the function `foo`, instead
of just `(float bar)`.

In addition, a similar test case to what was originally
reported in #2208 is added.

Something like

id id2("){ ... }");

is no longer wrongly recognized as a "function"

id id2(") {
  ...
}
");

As the difference in the tests shows, this has the unfortunate side
effect that we no longer highlight something like

int f(param="default");

as a function declaration, but it is hard to imagine another way to
fix this (cf. “most vexing parse” problem).

Fixes #2207

Co-authored-by: Jean Abou Samra <jean@abou-samra.fr>

This commit adds a new url field to a lexer, which can be used to link
to the language website, instead of relying on having the link in either
languages.rst or the docstring of the lexer. Additionally, it changes the
languages.rst file to auto-generate the list of lexers from the actual
source code, using the provided URL.

* Lex identifiers after `case` as constants

Add a state for marking identifiers preceded by  a `case` keyword as
constants.

Additionally, refactor the `label` rule to no longer permit a `case`
keyword before a label.

Consequentially, identifiers after a `case` keyword (like `foo` in
`case foo:`) are no longer wrongly lexed as `Name.Label`, but as
`Name.Constant`.

In addition, this fixes #2076, as multiple `case` keywords in one
line are lexed the same.

* Add test for multiple `case` keywords in one line

* Fix existing tests

* Lex `::` as Operator and not Name.Constant

After a `case`, when lexing a namespaced name, like `foo::bar`, lex the
namespace operator `::` as Operator, and not Name.Constant.

* Regenerate tokens

* Remove unused variable

This variable is unused since the first commit to this file
in the commit history, so it's probably safe to remove it.

* Lex identifier as label only if it's at line start

* Stop matching identifiers that begin with a digit

This is so we don't match numbers as labels with the new labels rule.

* Add label tests

* Fix existing tests