[Draft]Create `AsymmetricMetric`

[alex-hse-repository]

🚀 Feature Request

Create metric, which use the custom penalty for over- and under- forecasting. This might be useful when the over- and under- forecasting is harmful differently.

Proposal

• In file metrics/metric.py create class:

class AsymmetricMetric(Metric)

• Constructor should have the following interface:

def __init__(
    self,
    mode: str = MetricAggregationMode.per_segment,
    asymmetric_threshold: float=0.0,
    under_estimation_penalty:  Callable[[ArrayLike, ArrayLike], float],
    over_estimation_penalty: Callable[[ArrayLike, ArrayLike], float],
   **kwargs
):

Test cases

No response

Additional context

No response