aggregate over lead

[aaronspring]

Is your feature request related to a problem? Please describe.
I would like to verify total precipitation, i.e. precip pr aggregated over lead. I can easily create aggregated forecasts, but the verification data wont aggregate.

Describe the solution you'd like
add forecast.lead.attrs['aggregate'] = 'cumsum'?

Describe alternatives you've considered
see s2s-ai-challenge with xskillscore only

[aaronspring]

Two ways presented in #630:

• use HindcastEnsemble.smooth(dict(lead=14), how='sum') for 14D accumulation of precip
• use rolling sum/mean to get observations as biweekly timeseries for each day

However, this does not allow to verify skill over daily tp skill

[aaronspring]

climpred.HindcastEnsemble Loops over lead of initialised and then matches obs time. Therefore aggregation over lead, I.e. using precipitation rate pr to predict precip amount tp total precipitation, is currently not possible.