Plan for async data frames.md

This is my plan to support asynchronous (unresolved) dataframes.

I want to be able to do something like this:

dataForge.fromMongoDB()
    .head(5)                // This should be passed through to the query.
    .resolve()
    .then(dataFrame => {
        // Now have a data frame with 
    })

Or this:

dataForge.fromMongoDB(inUrl)
    .select(row => transform(row))
    .toMongoDB(outUrl)
    .then(() => {
        // Completed!!
    })

Or

dataForge.fromMongoDB()
    .toArray()
    .then(array => {
        // Baking operations normally in Data-Forge should return a promise for an async dataframe.
    })        

This requires new classes: AsyncDataFrame and AsyncSeries.

Then I can have overriden functions.

Normal functions, eg 'select' should just come from the base class.

These need to be retro fitted to build some kind of AST.