lalpulsar_ComputePSD: set sensible default options for spectral averaging

lalpulsar_ComputePSD has options for controlling how power spectra are averaged, e.g. what mathematical average to apply over SFTs from different times and/or different detectors. These days the default values for these options are not sensible, but have be retained for many years to keep backward compatibility. That runs the risk, however, that people are using the wrong options to compute PSDs, particularly for computing sensitivity depths, so it's time to update the default options.

For the options which control the computation of averaged PSDs, I propose these default values:

• --PSDmthopSFTs=harmsum
• --PSDmthopIFOs=harmsum
• --normalizeByTotalNumSFTs=true

These are the correct options for producing averaged PSDs over all SFTs (irrespective of time/detector) for computing sensitivity depths. A number of LALSuite and OctApps codes rely on the this convention.

Furthermore, I propose renaming the option normalizeByTotalNumSFTs to PSDnormByTotalNumSFTs. This option is only applied to computing PSDs and not normalised SFT powers (see below), so using the PSD prefix will make it clearer how this option is used.

lalpulsar_ComputePSD also has the option to compute normalised SFT powers (i.e. where the PSD is divided by its running median). This is turned on by --outputNormSFT and controlled by --nSFTmthopSFTs and --nSFTmthopIFOs. It is proposed to have no default values for the latter 2 options, but instead make them mandatory if --outputNormSFT=true, so the user will have to think about what operations they want to use.

cc @david-keitel @john-whelan @evan-goetz for comments.