Add LinearRegressionComputeController

[pockerman]

The LinearRegressionComputeController is responsible for fitting a linear model with a given data set

• GET:

1. Should allow the user to select the data set to be used for the fitting

2. From the specified data set the user should be able to:
   1. Select which columns will form the features of the model (at least on column is needed)
   2. Select the polynomial order of the feature (default should be 1)
   3. Select which column will be used for the labels. If the column selected is in the column set from 1 above an error should be generated

            4. Select the error function
                 1. Mean squared error
                  2. Sum squared error
           5.  Select regularization method
                  1. Lasso
                  2. Ridge
                  3. ElasticNet
            6. Select the optimization method
                  1. Gradient Descent
                  2. Newton solver
   

• POST: Accepts the filled user form and creates the task(s) for the computation

It redirects to a page where the computation completion can be queried. Thus we should be able to distinguish between computing tasks. e.g. use some sort of name mangling

Consider: We could write to a DB when a calculation is started and finished a poller can then query this and once the computation is finished serve the results somehow

The result is should be a set of coefficients describing the model.

Consider: The user should be able to save in the DB the computed model for future use under a name