logger in mdpow prints out correlation time (g/2) as the statistical inefficiency (g)

[VOD555]

The logger in the mdpow outputs the correlation time (g/2) but not the statistical inefficiency (g). And as the one calculated in mdpow is based on the length output of the subsampled dataframe, there is small difference between the mdpow one and alchemlyb one.

As currently alchemlyb has a logger output for the SI, I think we can remove the mdpow logger output for g.