Standard for documentation types

Tell us about it

Documentation uses numpydoc and is linted using pydocstyle. However, the format of the type is not checked. Specifically, densityplot.py now uses proper type hinting, while everywhere else uses a less rigorous style emulating other pydata projects (int, optional, instead of Optional[int], for example)

I am most in favor of consistency across the code base, which would suggest changing densityplot.py. I would also prefer type hinting everywhere to our current standard.

Specifically: I think we should manually urge new contributions to match current documentation practices, but I would also accept a pull request that converts a reasonable portion of the codebase to the new style.

Thoughts on implementation

My feelings on this are weak enough that I do not think it should be in CONTRIBUTING.md, since every requirement there feels like a hurdle to engagement. If we ended up in a place with type hinting everywhere, then I would be in favor of adding it.

[OriolAbril]

mypy is to be used in the code itself, not in the docstrings. See example below.

def from_pymc3(
    trace: MultiTrace, prior: Optional[Dict[str, np.ndarray]]=None, ...
) -> InferenceData:
    """
    Converts ...

    Parameters:
    trace: pymc3.MultiTrace
        description
    prior: dict of {str: np.ndarray}, optional
        description
    ...
    """
    code
    ...

[percygautam]

@OriolAbril So, we have to extend the contribution guide about new links suggested by @canyon289 . This can also be added to the new PR template being added. Also, are you suggesting to change all docstrings to mypy format?