[ENH] create_pipeline function

[samihamdan]

A couple of issues (e.g. #112 #113 ) are based on the idea of using a julearn model/pipeline in other function then the run_cross_validation. In some situations I would even like to use a model in a custom/manual cv loop.

For these situations it will be helpful to have a public function which uses the same inputs as run_cross_validation, but returns only a not trained pipeline which can be used however the user wants.

[fraimondo]

Was this issue addressed in #115?