Add notebook dependency to testing_py35.yml

[randyzwitch]

On master 8776841, the conda environment in notebook_py35.yml is broken, giving error messages similar to #101 and #117. I believe this is due to https://github.com/gpuopenanalytics/pygdf/blob/master/conda_environments/notebook_py35.yml#L84-L85 having outdated version libraries.

For maintainability, it seems like bundling Jupyter notebook within the pygdf_dev environment makes more sense, and then two .yml files don't need to be maintained. I labeled this PR as [WIP] because I didn't want to delete the notebook_py35.yml file in case someone thinks it should stay, but if everyone agrees I can push another commit to remove notebook_py35.yml.

[mflaxman10]

Got burned by these old dependencies myself before noticing this PR. I second the motion for something like this.

[sklam]

The notebook_py35.yml has thrift package for the mapd integration notebooks. So, i think it is still needed. But, we will want to move/update them since mapd has changed since those notebooks were added. FYI, details info about setting up the mapd notebooks are at https://github.com/gpuopenanalytics/pygdf/blob/master/notebooks/README.md.

[randyzwitch]

I think the overarching point is whether there needs to be two separate environments. Currently, the notebook environment doesn't work (nor do the notebooks, fwiw). So making one common conda environment would probably help in that regard.

[sklam]

Right, a single env is good. There are just some housekeeping task before the notebook env can be removed. The notebook env is really for the mapd notebooks and they should be updated and moved into somewhere else.

As for this patch, I am happy to merge it. Remove the "WIP" to indicate that it's ready and I will merge it.

[randyzwitch]

Ok, lets merge this for now, the notebooks can be handled separately. I'm happy to work on the MapD notebooks examples, as I joined MapD last week.