Logging to MLflow in pipeline.eval()

[tstadel]

What?
As a user I want to store my EvaluationResult in MLflow.

How?

• We log EvaluationResults (metrics + eval dataframes) of pipeline.eval() to MLflow.
• Each eval() call creates a run in MLflow.
• Other logging engines should be supported. Default is None, MLflow is currently the only implementation.
• See https://github.com/deepset-ai/FARM/blob/2fc801d1d0c973cb36bc867907d4250d1084493b/farm/utils.py#L148 for code reuse
• We need to pass tracking_url, experiment_name and experiment_run_name (might be coupled into a ExperimentRun object)