Splitting fitting/prediction and plotting?

[dtolpin]

Hi,

can you consider splitting the core functionality and the plotting? I (and probably others) want to run pydlm in the server environment where plotting is not required. In some environments the space is tight and installing matplotlib may be a problem. In particular I am going to run pydlm on AWS Lambda, where I need to prepare a stripped down version of pydlm with plotting (or at least the matplotlib dependency) removed.

It would be easier to just have this in two separate packages.