enh: add doctests to travis (#2)

[sjsrey]

No description provided.

[ljwolf]

I believe the shapley_ext.py errors are coming from the last line, where the for method in __all__ copies over all the shapely docstrings carte blanche. This is fine to merge as is.

But, consider:

1. the shapely_ext stuff isn't unit-tested (in libpysal/cg/tests)
2. as written, they won't be doctested anymore (they borrow shapely's doctests, so have no control over their passing)
3. we tend to recommend folks go straight to geopandas/shapely for data read/write and GIS in our workshops
4. they're only depended on by the wkb reader for testing, which, again shapely.wkb.loads should be recommended.

Like at the Friday call, I really think we should delete libpysal/cg/shapely_ext.py rather than baking it in to be ignored in testing. I think we should remove functionality if it's not tested, not used, and not recommended to be used.

[sjsrey]

I agree with the idea of removing libpysal/cg/shapely_ext.py.
Should we deprecate this first, or just remove? Either way, it looks like, in addition to being used by the wkb reader, it is also used in cg/ops?

[ljwolf]

good check; that is a soft dependency in cg/ops.
It's probably a good idea to deprecate... I know my tendency is to want to just remove immediately, but that's not fair to others using it.

Let's merge this. I'll try to get a PR in adding deprecation warnings to the shapely_ext or make sure they are unittested.