Draft: Fix uncertainty calculation sampling GPR

This fixes a potential bug in the uncertainty calculation when sampling the GPR. Since the GPR was computed on the real and imaginary terms, we should sample using real and imaginary. It had been spotted that at high frequencies we would sometimes get random 180 degree phase jumps.

Quantiles are computed on the real and imaginary terms first before converting to magnitude and phase. This should make the calculations more robust.

This also updates the interpolation due to deprecation and to make sure the interpolation on complex valued terms are handled separately.

Make consistent the use of logspace in pydarm (remove one use of geomspace in cmd)

Fix a small bug when previous_model_response_curve_file is not provided in the uncertainty configuration.

An example output (at least so far):

Closes #215