ENH: Implement legend (colorbar) for noncategorical dataframe plots

[ianalis]

This PR implements the legend option for noncategorical dataframes which, up to now, returns a NotImplementedError. Since there is no easy method of accessing the legend of an axes that I'm aware of, plot_dataframe() will also return the colorbar aside from the axes. Additional tests and appropriate changes in the documentation were also added.

[ianalis]

Travis CI build failed because of ImageComparisonFailure in my new tests. Is there a way to get the generated images in the build? Obviously, the tests pass on my machine.

[jankatins]

We had similar problems in ggplot and this went away when we made some changes to the test setup:

https://github.com/yhat/ggplot/blob/master/ggplot/tests/__init__.py#L230

The problems are IMO because of difference between windows and linux/travis...

(we also copied the testing functions to get them working in earlier verisons of mpl, which didn't allow for tests outside of mpl test directory)

[coveralls]

Coverage increased (+4.61%) when pulling f3b12a4 on ianalis:noncategorical-legend into cf4eb4d on geopandas:master.

[coveralls]

Coverage increased (+4.61%) when pulling dcb7f60 on ianalis:noncategorical-legend into cf4eb4d on geopandas:master.

[coveralls]

Coverage increased (+4.61%) when pulling 43333f3 on ianalis:noncategorical-legend into cf4eb4d on geopandas:master.

[ianalis]

Thanks for the tip @JanSchulz. I adapted the settings from ggplot and explicitly set the sans-serif font, as well. I printed the base64-encoded failing images to download the images from travis but I removed that part when I rebased my PR.

[coveralls]

Coverage increased (+4.64%) when pulling 700c330 on ianalis:noncategorical-legend into cf4eb4d on geopandas:master.

[nickeubank]

@ianalis This looks like a really nice addition -- any chance we can get this back to testing clean? We're hoping to (finally) get release 0.2 in the next month or two, and this would be a great addition.

[IamJeffG]

#267 includes a fix for this issue.