Test SMOGN end-to-end via API launch

[stefangary]

After the SMOGN pull request, now I want to test that SMOGN works (or is turned off) via API launch options. Through using the API launch, we can also automate the process of running SMOGN to check for the mean and variability of the results of applying SMOGN and also test whether the SEED enforces reproducible results.

[stefangary]

There are currently two issues with the SMOGN integration (potentially related):

1. SMOGN fails occasionally (i.e. one of 10 SuperLearners will fail during the SMOGN process) due to presence of NaN.
2. There are dataframe copy of view warnings - need to take a closer look at Pandas DF management. Maybe use .iloc().