To avoid the crash of ModuleAnalyzer from invalid type comments, this
start to retry parsing without type_comments=False when `ast.parse()`
raises SyntaxError.

.. note::

   $ find sphinx tests LICENSE doc/conf.py -type f -exec sed -i '' -e 's/2007\-20../2007-2021/' {} \;
   $ git co sphinx/locale/**/*.js sphinx/templates/epub3/mimetype

On the renaming ModuleAnalyzer.parse(), I overlooked renaming a flag.

The word "analyze" is much appropriate for "ModuleAnalyzer" instead
of "parse".

Sort imports with isort

Keep imports alphabetically sorted and their order homogeneous across
Python source files.

The isort project has more feature and is more active than the
flake8-import-order plugin.

Most issues caught were simply import ordering from the same module.
Where imports were purposefully placed out of order, tag with
isort:skip.

* The result of ModuleAnalyzer.parse() is not cached
* autodoc tries to search overloaded constructor methods to the root
  class even if a definition found

Developers can write number literals in several ways. For example,
decimal (1234), hexadecimal (0x1234), octal decimal (0o1234) and so on.
But, AST module don't mind how the numbers written in the code. As a
result, ast.unparse() could not reproduce the original form of number
literals.

This allows to construct number literals as possible using original
source code.

Note: This is only available in Python 3.8+.

This implements tuple literal support to sphinx.pycode.ast.unparse().

It is worthy to keep the order of analyzer.attr_docs to generate
document in reproducible. So this uses OrderedDict explicitly to do
that. It also helps python3.5 environment.