Fixes to docs, eliminate scikits.timeseries docs dep, sphinx 1.2, update copyright years

[ghost]

closes #5836
closes #5899, following @jorisvandenbossche's advice

[ghost]

Note the autosummary removed was actually a template, not duplicated code, but
it's role was to filter out private methods and I didn't see a problem since api.rst
explicitly states the methods refernced in the generated documentation, and it broke
recent sphinx 1.2.

If I missed something there, speak up.

[jorisvandenbossche]

@y-p Yes, I am afraid you missed something. Can you revert the commit which removes autosummary template?

The template causes the error with Sphinx 1.2 as explained in #5899, but it is still useful. But maybe my wording there was not very clear. Sorry for that. When I said that removing the template solves the error, I didn't mean that the template should be removed ... It was just to know the error has something to do with the template (to further investigate it) and for people who really want to build the docs with 1.2.
The function of the template is to automatically create docstring pages for all methods of DataFrame, Series etc (because not all of them are added manually to the api.rst page) so linking to them from the docs works and they can be found with google. See #5160 for details.

[ghost]

oh shoot, I see. How are those pages reachable to users, do they have to guess a direct link?

[ghost]

We can revert, but iiuc Up until 3 months ago we didn't have this and it breaks the stable version
of sphinx? perhaps this should be done another way.

[jorisvandenbossche]

These pages are directly reachable via eg http://pandas.pydata.org/pandas-docs/dev/generated/pandas.DataFrame.html#pandas.DataFrame. And you can also get there eg by searching google.

And you can also say new Sphinx breaks our doc building :-) I just mean, at the time I added it, it worked with the current stable version of Sphinx.

[ghost]

Hey, that is nice.

I guess sphinx did, but given that pip install sphinx now can't build our docs do you still think
we should do this? if so, I'll revert.

[jorisvandenbossche]

Of course I think we should solve it, and ideally as fast as possible. But I think keeping the dev docs on the website up to date (as long as the doc building there still happens with 1.1.3) is more important than having it buildable by users/developers for the short term. As they should normally be able to install 1.1.3 manually if they are explained to do. We should then add a note about that in the guidelines to build the docs for now. What do you think?

And if you think of other ways to obtain similar results, it would certainly also be good. I just copied the approach of numpy and scipy (they have also such a template class.rst, for this reason at the issue I pinged pv to ask if they have the same problem).

[jorisvandenbossche]

BTW, there is something wrong with the docs building, as the dev docs are last built on Jan 3, so ten days ago.

[ghost]

The docs on the website aren't building currently, but that's a whole nother thing.
My preference is for users that try to build pandas to have a good experience and
I'm willing to sacrifice some recent aesthetic addition for that until we find a way to do both.

But - I don't mind yielding here. Reverted.

[jorisvandenbossche]

See my comment on #5899, regression is already fixed in sphinx development version. But of course, that does not yet solve the issue for users/developers who want to build the docs now.

[ghost]

Yeah, I would have preferred tabling shelving this until pypi version is fixed but given that
we know this will resolve itself in time, it's ok by me.