Speed up Similar Cases page load time

[zyteka]

The similar cases page can have a very long load time for larger datasets (~22 seconds for brakepad from my experience). Per #82 , this load time also prevents over pages from loading. sibyl-dev/pyreal#503 and sibyl-dev/pyreal#505 may fix this, but if not we should investigate ways to prefit the neighbors or reduce the dataset used for this page specifically.

[zyteka]

Profiling has shown that fit/transform of the similiar examples explainer is actually very fast right now. Most of the runtime comes from having the run the MultiTypeImputer to transform the training dataset before getting similiar examples. This would normally be handled during fitting (for other explainers), but must be handled at runtime for the similar examples explainer or the explainer would be too large to fit in the dataset.

On the bright side, the new similar examples implementation with the current brakepad dataset is small enough to fit in the mongodb database. Its not clear how future-proof this is, but does speed up this page for the time being. The long-term solution is probably to store the model-ready data in the database