Ability to save pipelines using pickle

[dsherry]

Background
The pipeline save and load methods currently use cloudpickle instead of python pickle.

We closed #1400 a couple months ago. But this led to a bug (#1912) where our code was setting class attributes which apply across all instances of our template classes.

Plan
The short-term resolution for #1912 is to revert the change for #1400, so that get_pipelines returns a pipeline instance which is not pickle-able using python pickle but also does not have the buggy side-effect behavior when it comes to the class attributes like name and component_graph

Once that's done, this issue tracks figuring out how we can modify our approach to support python pickle of our pipelines, and filing issues to execute that plan.

[angela97lin]

The issue at hand is that make_pipeline returns a dynamically-generated class and pickle is not able to pickle dynamic classes.

Options:

• Be explicit about not supporting pickle and requiring users to use cloudpickle if they want to save their pipelines.
• Right now, we need this dynamically-generated class because we expect a pipeline class to have specific attributes (custom name. component graph, custom hyperparameters). We could move from storing those as class attributes to instance attributes. These were added in Pipeline API #345 to facilitate the ease of creating new pipelines, but it could be worth revisiting again.

[dsherry]

@chukarsten suggested looking at defining reduce on our pipelines. I think that could be the way to go to make pipeline classes serializable with pickle. https://stackoverflow.com/questions/11658511/pickling-dynamically-generated-classes

However that does still leave us with another challenge. Currently we define the pipeline classes used in automl at runtime, and as @angela97lin said above, the original issue was that those definitions weren't being found/saved during the pickling, and that issue still needs a solution.

[dsherry]

We did some digging this morning and listed options.

[angela97lin]

Next steps: open a spike PR, see what difficulties we encounter with updating our pipeline API.