Check snowfall workflow

[LouiseABowler]

Similar to #3 - let's check that the script for the snowfall dataset, whistler_ICML18.py generates the values needed for Table 1. There isn't an associated paper_pictures... script for this dataset as the results are only used for the table.

• Check paths to data.
• Output error measures for Table 1.

[LouiseABowler]

The snowfall script now outputs the values needed for Table 1: the MSE, NLL and their associated 95% errors.

It takes a lot longer to generate results than the Nile example does, so it would great if we can get some of the unused options at the top of the file back into play - e.g. we could load a previously saved set of results if run_detectors is false. We should be able to extract MSE and NLL directly from the saved file, but that isn't working right now - let's get some tests in first and then I'll get started on checking the save/load process for Evaluation_tool.