Climlab

This release has some under-the-hood improvements to the Fortran builds that will let us build properly for a wider variety of platforms (including Windows / Python 3).

With thanks to xoviat

This is an experimental release with an improved build scheme. This release is really just to test the automated build procedure on conda-forge.

This release provides

- the [Emanuel convection scheme](https://emanuel.mit.edu/problem-convective-moistening), so we can now set up and run a comprehensive single- or multi-column radiative-convective model with interactive water vapor
- support for asynchronous coupling of time-dependent components
- significant optimization of some climlab internals to speed up the runtime

No backwards-incompatible changes have been introduced.

This release brings basic integration with [xarray](http://xarray.pydata.org/en/stable/).

There are now `to_xarray()` methods for automated translation of climlab model variables into `xarray.DataArray` and `xarray.Dataset` objects. See [the new docs](http://climlab.readthedocs.io/en/latest/xarray.html).

This should simplify analysis and plotting of model output.

Deeper integration with `xarray` is planned for later releases.

With this release, climlab is now fully Python 3 compatible!

Other goodies include:
- improvements to the documentation
- a recipe for building climlab with `conda-build` is now included
- a few bug fixes and minor enhancements

This release fixes some issues with the source distribution that are causing problems building on conda-forge.

No changes to climlab functionality.