passing parameters to metrics

[s6sebusc]

Description

Parameters for individual metrics are added in a new optional entry metric_parameters in the eval config, for example

evaluation:
  metrics  : ["fbi", "rmse"]
  metric_parameters:
    fbi:
      thresh: 0.001
  regions: ["nhem"]
  summary_plots : true
... 

When a metric is loaded or computed, corresponding parameters are taken from metric_parameters.

New JSON files for scores now have a top level key "scores", below which is a list of results for the same score with different parameter settings, with each list entry following the same format as before. Old files with no "scores" key are treated appropriately.

Upon loading a score, the reader goes through all elements of the JSON file in search of the right parameter settings. When saving scores, metric_list_to_json now checks whether the output file is already present. If it is, it goes through the file in search of the right parameters settings. If it finds them (possible when for example fsteps was changed), it replaces that entry with the new values. If the right parameters are not found, the new results are appended at the end of the JSON file.

Issue Number

Closes #1475

[s6sebusc]

I have not yet looked at the map plots or mlflow.