on selection of multiple points, show how they are different from all other points

[micahstubbs]

depends on #19

design:

1) lasso select some points 2) create new property for every point on the plot:

• 1 if the point is in the lasso selection
• 0 if the point is not in the lasso selection 3) create a new frame in h2o-3 that contains the data from all the points on the plot, with the new lassoSelected binary property 4) train some model on that new frame
• start with a random forest model 5) poll h2o-3 to find out when training is complete 6) when training is complete, get model metrics from h2o-3 7) parse out training variable importance data from model metrics 8) show top n important variables for lasso selection
• in a tooltip?
• in a linked view?

[micahstubbs]

lasso selection could be a group of points like this