Create examples and improve documentation to *explore partitioning first, learn models second*

[cchriste]

This video demonstrates the issue:

• using knn=15 for the June2020 Cantilever Beam dataset results in a single partition when computing the M-S for the Compliance QoI field.
• If knn is reduced to 5, there are 9 partitions, much more interesting for a user to explore and identify different trends in the data.
• this is why we provide a data processing pipeline and user guidance (i.e., documentation and examples) showing the process

We need better examples as our own current example datasets all use knn=15 and don't demonstrate the flexibility of the design Space eXplorer.