Record evaluation outputs

[roman807]

Record evaluation outputs and save them under the following folder structure:

|-results
  |-<run_id>   # based on timestamp + hash based on config, one "eva fit/predict" call corresponds to one run
    |-artifacts   # outputs from TensorBoardLogger
       |- checkpoints
       |- configs.yaml
       |- tensorboard.ervents
       |- metrics.csv
    |-<run-config>.yaml   # copy of run config
    |-results.json   # json with start-time, end-time, metrics of all val/test results

[roman807]

duplicate #158