Divide by zero still occurs with add_espilson = True for PearsonCorrelation and Rsquared

[kvanwerkhoven]

setting add_epsilon = True for these functions has no effect. Still getting divide by zero warnings. Looks like both use np.corrcoeff (ignores epsilon). Likely need to calculate directly instead when epsilon=True.

[samlamont]

Implement pandas or scipy spearman instead numpy

Pandas:
p.corr(s, method='spearman')

[samland1116]

Closed by #562