[MLOps 1.5] Autogenerate data logging configs from the registry

[dbogunowicz]

Feature Preview:

string_result = data_logging_config_from_predefined(
        group_names="image_classification",
        frequency=2
        loggers={"some_logger": {"arg_1": 10}},
    )

outputs:

loggers:
  python: null

data_logging:
  pipeline_outputs.labels:
  - func: predicted_classes
    frequency: 1
  - func: predicted_top_score
    frequency: 1
  pipeline_inputs.images:
  - func: image_shape
    frequency: 1
  - func: mean_pixels_per_channel
    frequency: 1
  - func: std_pixels_per_channel
    frequency: 1
  - func: fraction_zeros
    frequency: 1

Description

Introduces a new helper function, data_logging_config_from_predefined, that generates data logging configurations using metric group names and optional arguments. The function enables developers to quickly generate examples of data logging configurations, as demonstrated in the examples/data-logging-configs directory.

This PR includes comprehensive testing and features a test that, when run with the appropriate environment variable, can regenerate a fresh set of examples. This functionality is useful for maintaining up-to-date configurations as the log structure evolves over time. :

GENERATE_CONFIGS=1 pytest tests/deepsparse/loggers/metric_functions/helpers/test_data_config_from_predefined.py 

[corey-nm]

the yamls look good to me!