Giddy

This release marked a transition from supporting python 3.5 and 3.6 to python 3.6 and 3.7

Full release details are reported in [CHANGELOG.md](https://github.com/pysal/giddy/blob/master/CHANGELOG.md).

This release features the addition of two rank-based Markov classes proposed in [S. J. Rey. Rank-based Markov chains for regional income distribution dynamics. Journal of Geographical Systems, 16(2):115–137, 2014.](https://link.springer.com/article/10.1007/s10109-013-0189-0):
* [Full Rank Markov](https://github.com/pysal/giddy/blob/master/giddy/markov.py#L1728)
* [Geographic Rank Markov](https://github.com/pysal/giddy/blob/master/giddy/markov.py#L1867)

[A Jupyter notebook](https://github.com/pysal/giddy/blob/master/notebooks/Rank_Markov.ipynb) was prepared to illustrate the application of these two classes to reveal interesting spatiotemporal patterns of income distribution dynamics in the United States 1929-2009.

Full release details are reported in [CHANGELOG.md](https://github.com/pysal/giddy/blob/master/CHANGELOG.md).

This release does not add any new functionality to `giddy`, but instead features api changes in `giddy` and its `pysal` submodule dependencies `mapclassify`, `libpysal`, and `esda`. More specifically, the `giddy.api` module which was originally designed to facilitate a smoother transition from the old metapackage pysal to the refactored submodule structure (see [here](http://pysal.org/about.html#migrating-to-pysal-2-0) for details) was removed as we are moving away from the refactoring stage and looking at the future development.

This release features:
* a more flexible specification for the [spatial Markov chains model](https://github.com/pysal/giddy/blob/master/giddy/markov.py#L169). More specifically, for continuous time series input:
* both the numbers of classifications for input continuous time series (k) and spatial lags (m) can be specified and allowed to be different
* user-defined classifications (cutoffs) for input continuous time series and spatial lags are allowed
* new visualization tools:
* relies on [pysal/splot](https://github.com/pysal/splot) for [visualizing dynamic LISA related statistics](https://github.com/pysal/giddy/blob/master/giddy/directional.py#L322)
* the launch of giddy [documentation website](http://giddy.readthedocs.io/)

Full details are reported in [CHANGELOG.txt](https://github.com/pysal/giddy/blob/master/CHANGELOG.txt).

This release is the first tagged release of giddy on Github.

Starting from this release, giddy supports python 3.5 and 3.6 only.

This release also features categorical spatial Markov which enables spatial Markov (class [Spatial_Markov](https://github.com/pysal/giddy/blob/master/giddy/markov.py#L179)) to be applied to categorical time series such as land use and land cover change, as well as neighborhood change. Here, the spatial lag (utilizing function [lag_categorical](https://github.com/pysal/libpysal/blob/master/libpysal/weights/spatial_lag.py#L88) in [libpysal](https://github.com/pysal/libpysal)) is defined as the most common category among neighbors.

This is a major release incorporating migration to python 3+ and new features. Full details are reported in [CHANGELOG.txt](https://github.com/pysal/giddy/blob/master/CHANGELOG.txt).