[ENH/API] have a feature computing function that returns the covariance data frame

[dengemann]

The idea is to hide the construction of cov data frames from the user.

Say:

cov_df1 = compute_coffeine(inst=raw1, method='frequency_bands', method_params=dict(bands='ipeg'))

cov_df2 = compute_coffeine(inst=raw2, method='frequency_bands', method_params=dict(bands='ipeg'))

cov_df = pd.concat([cov_df1, cov_df2])