Failure to parse docstrings where a section contains header-like content

[aaronfng]

numpydoc.docscrape.NumpyDocString in version 1.9.0 fails to parse docstrings if there is content within a section that resembles a section header, e.g. the "Parameters\n----------" contains " Parameters\n----------"

Example reproducer: a function that takes a function as an argument and returnsanother function. The following docstring explains the returned function:

def wrapper(func):
    '''
    Wrap a simple function so that it takes a message string that gets printed to stdout when invoked.

    Parameters
    ----------
    func: function
        Function to be wrapped. This can be any arbitrary synchronous function.

    Returns
    -------
    wrapped_function: function
        The returned function has the following signature::

            wrapped_function(
                message,
                *args,
                *kwargs,
            )
                \'\'\'
                Parameters
                ----------
                message: str
                A custom message to be printed when the wrapped function is called.
                *args
                    Additional arguments that are passed into the original function.
                **kwargs
                    Extra arguments to the original function.
                \'\'\'
    '''

However, when using numpydoc.docscrape.NumpyDocString to parse this docstring, for example:

parsed = NumpyDocString(<example docstring above>)

...an exception is raised claiming to have duplicate Parameters section:

ValueError: The section `Parameters` appears twice in  
Wrap a simple function so that it takes a message string that gets printed to stdout when invoked.

The desired behavior is to treat the nested Parameters "section" as regular content and not an actual section.