PSD issues for PE

I was checking on some recent PE jobs and saw a few issues with the PSDs being used.

• PSDs impacted by the low-pass filter. Low-pass filtering of the data can lead to power suppression at high frequencies. When this is the case, we need to reduce the upper-frequency limit to below the affected region. I think this is particularly important when using PSDs from SPIIR. I copied in some plots below of what this issue looks like (from G963601), the other thing we see is the evidence is estimated as extremely large O(10000-1000000).

• PSDs estimate lines as less significant than PE analysis. I suspect this is because shorter segments are being used to estimate the PSD than the analysis segment (PSD resolution is 0.25 Hz, so I'm assuming 4s FFTs are used). I'm not sure exactly what the impact of this will be on the results, but will likely lead to the SNR being overestimated/posteriors being over-constrained.

For reference, these are the equivalent plots for GW150914 using the production PSDs