dynamic pipeline loading

[yaythomas]

refactor to allow dynamic plug-in approach to discovering and loading pipelines.

def load_and_run_pipeline(
        pipeline_name,
        pipeline_context_input=None,
        working_dir=None,
        context=None,
        parse_input=True,
        loader=None
):
    # the default loader is the file loader.
    if not loader:
        loader = 'pypyr.pipeloaders.fileloader'

     # get loader dynamically
     loader_module = pypyr.moduleloader.get_module(loader)
     pipeline_definition = loader_module.get_pipeline_definition(...)

Refactor the current get_pipeline_definition() to a new pypyr.pipeloaders.fileloader

Add an optional loader param to pypyr.steps.pype step. this needs to be completely backwards compatible. In get_arguments:
loader = pype.get('loader', None)
and add to the return tuple.

Custom loaders need to adhere to the signature get_pipeline_definition(pipeline_name, working_dir), returning a dict-y pipeline definition.