Windows doesn't support symlinks out of the box, and there doesn't
seem to be any use of this symlink, so let's remove it.

fixes #1548

fixes #1223

* Add a check for CR/LF in files.

This can occur when checking out things on Windows, and it breaks the
tarball. This adds a script to check for the presence of CR/LF which
exits early if anything gets found.

* Improve error checking.

* Include the external folder and check that.
* Include .bashcomp files.
* Use the correct CR/LF on the checker itself.

* Address review feedback.

* Remove || true
* Fix docs
* Print the first offending file name

* Add analyze_text to make make check happy.

This also fixes a few small bugs:

* Slash uses *.sla as the file ending, not *.sl
* IDL has endelse, not elseelse

* Improve various analyse_text methods.

* Improve various analyse_text methods.

* Make Perl less confident in presence of :=.
* Improve brainfuck check to not parse the whole input.
* Improve Unicon by matching \self, /self
* Fix Ezhil not matching against the input text

* Simplify Modula2::analyse_text.

* Add regexlint to tox so linting can be done during development

Tested on Windows.

Closes #1556

* Make regex linting depend on pygments/regexlint@master

* Fix LatexEmbeddedLexer to not call the nested tokenizer piecewise

* Reuse the existing do_insertions function

* Add a test for the `LatexEmbeddedLexer`

@belett noticed that VALUES was missing [1]; I found the other ones by
running the following snippet on the SPARQL 1.1 Query Language spec:

    new Set(Array.from(document.querySelectorAll('.grammarTable'))
        .reduce((text, elem) => text + elem.textContent)
        .match(/'[a-z0-9-_ ]*'/ig))

I don’t know why a few keywords were missing; the docstring linked to
the SPARQL 1.0 Query Language spec (also fixed here), but the lexer
already contained other tokens which were only added in SPARQL 1.1, such
as the aggregate functions (MIN, MAX etc.), which have already been in
Pygments since the initial commit of the current history (6ded9db).

[1]: https://phabricator.wikimedia.org/T264175

Co-authored-by: Matthäus G. Chajdas <Anteru@users.noreply.github.com>

* TNTLexer: Don't crash on unexpected EOL

Catch IndexErrors in each line and error the rest of the line,
leaving whatever tokens were found.

* Write and pass tests for Typographic Number Theory

pygments/lexers/tnt.py:
* Fix indentation on import
* Fix: TNTLexer.cur is class-level reference if not initialized
  in get_tokens_unprocessed, so init it in __init__ too
* Fix: Fantasy markers are not allowed as components of other formulas,
  so have a dedicated check for them in the body of get_tokens_unprocessed
  which disables the normal formula handling if present
* Clarify TNTLexer.lineno docstring
* Attempt to discard tokens before an IndexError

+tests/test_tnt.py:
* Test every method, and test both +ve and -ve matches for most
* Lexer fixture is test-level to reinitialize cur clean each time
* Don't test actual get_tokens_unprocessed method (besides for fantasy markers)
  because the text testing is left to examplefiles

AUTHORS:
+ Add myself to credits :)

* Add a TNT test just to make sure no crashes

This removes the top/bottom padding changes, and only keeps left/right
padding, in the hope that this does not break all Sphinx themes.

Update CHANGES, bump version.