Negative values in PSD lead to negative relative bandpower about yasa

[raphaelvallat]

Added warning in b0251d3

[raphaelvallat]

Potential solutions:

1. Subtract the minimum value of the PSD to force the PSD to be all-positive. However, by doing so we lose the original "unit" of the PSD, and therefore the absolute power will be wrong (as well as the TotalAbsPower column when calculating the relative power). Alternatively, we could also use a min-max normalization to scale the PSD from 0 to 1, but it suffers from the same issues.

2. Clip the negative values: however, that doesn't sound optimal esp. if a large proportion of the PSD is below zero.

In any case, YASA should return a warning when negative values are present in the PSD.