Avg Up Month/Down Month may not be aggregating correctly

[memura]

I find that if I pass in both a strategy and benchmark to calculate metrics, where the benchmark returns are all positive (like a cash index) the avg_loss function (quantstats.reports.metrics l.1139) ends up removing all rows as NaNs are present in every row for the benchmark, arising from the line
quantstats.stats.avg_loss l.223: returns[returns < 0].dropna().mean()

The strategy and benchmark need to be treated separately in the _stats.avg_loss function when calling dropna() ... may also be happening for _stats.avg_win function and anything else with dropna() in it.