aggregate the result of DocumentStatistic

[ArneBinder]

This PR modifies the DocumentStatistic

• to return aggregated results instead of the lists of collected values. The default aggregation functions can be set via the class variable DEFAULT_AGGREGATION_FUNCTIONS: List[str] (the fucntion names are resolved against some inhouse methods (mean, median, std), or python builtin functions and, if this does not work, against the global namespace via utils.hydra.resolve_target).
• adds the parameter show_histogram: Iff True (default: False), show the histogram of the collected values with plotext on the console
• adds the parameter show_as_markdown: Iff True (default: False), show the result as markdown on the console (requires the tabulate package)
• adds the parameter title: this is used in the histogram and markdown table captions

This also adds the property current_split to the DocumentMetric which is available when processing a dataset dict.

Note: This is breaking with respect to #312 because before this PR, the DocumentStatistic returned a dict with list values, but now the values are aggregated (we do not label it as breaking because #312 was not yet part of a release). However, by using aggregation_functions=["list"] the previous behavior can be more or less restored (just the keys are a bit different).