Fix floating point precision issue

[ajithabhks]

Noticed that the auto powers having some very small imaginary values (order of 10^-12) in band averaging. It seems to be some rounding error due to numpy vectorization.

Possible solution would be manually make imaginary parts as zero using script (example):

self.bandavg_ds['exex'] = self.bandavg_ds['exex'].real + 0j

This is not a critical issue.