Create a Velox programming guide

[dcrankshaw]

Here's a preliminary outline. We probably don't need all of this ready for the first release.

• [x] Quick Start - should allow the user to get Velox up and running locally as simply as possible
• [x] Velox Overview - useful for context, but user doesn't need to read to get Velox working
  • What is velox? What problem does it solve? How does it fit with the rest of analytics?
  • Basic architecture
  • Models in velox
• [x] Getting started
  • Running Velox locally - an expanded version of the quick start
  • Configuration
  • Getting predictions
  • Making observations
• [ ] Examples overview - a brief explanation of what each of the examples do
• [x] Cluster Deployment - how to run multiple velox partitions on a cluster
• [x] Spark and Velox - how to do offline retraining via Spark
• [ ] Adding your own model
• [x] Contributing to Velox

[dcrankshaw]

We need to fix how models are added to Velox, so I am delaying a writeup of how to do it until the process is fixed. See #75 to track the model implementation fix.