Create a KeystoneML style guide

[tomerk]

Should contain information such as:

• What should be a Transformer vs. Estimator vs. LabelEstimator vs. FunctionNode vs. Evaluator vs. Loader vs. util method
• What packages to put things in
• (Till we figure something better out) ensure that Transformer RDD method implementations maintain each item & partitioning, e.g. functionally equivalent to rdd.map(x => apply(x))
• Using MatrixUtils.rowsToMatrix instead of DenseMatrix(x:_*)
• When to use Vector vs. DenseVector, when to use Matrix vs. DenseMatrix
• (eventually) when & how to use NumericTransformer

etc.
Also, we should decide and be consistent about whether nodes:

• are normal classes
• are case classes
• are normal classes w/ companion object constructor

[etrain]

Should also contain a link to the spark style guide (which is our coding standard) and something brief about the expectation of how docs are formatted and desired test coverage.