Add RollingOutlierCount primitive

[tamargrey]

Featuretools could include a RollingOutlierCount transform primitive that, like the other rolling primitives, takes in window_length and gap parameters and allows users to create features that indicate a level of volatility in their data over rolling windows.

class RollingOutlierCount(TransformPrimitive):
    name = "rolling_outlier_count"
    input_types = [ColumnSchema(logical_type=Datetime, semantic_tags={'time_index'}), ColumnSchema(semantic_tags={'numeric'})]
    return_type = ColumnSchema(logical_type=Double, semantic_tags={'numeric'})

    def __init__(self, window_length=3, gap=0, min_periods=0):
        self.window_length = window_length
        self.gap = gap
        self.min_periods = min_periods

    def get_function(self):
        def rolling_outlier_count(datetime, numeric):
            x = pd.Series(numeric.values, index=datetime.values)
            rolled_series = _roll_series_with_gap(x,
                                                  self.window_length,
                                                  gap=self.gap,
                                                  min_periods=self.min_periods)
            # get_outliers would be defined elsewhere, but it could use woodwork's box_plot_dict
            return rolled_series.apply(get_outliers).values
        return rolling_outlier_count