Remove dangerous defaults from METSignificance

The METSignificance tool needs to know the momentum resolutions of its input objects in order to do its calculation. These are provided by calibration tools from the various CP domains. If the user does not provide configured tools for this, METSignificance initializes its own with default configuration. The problem is that these configurations do not necessarily match what an analysis is using, so there is a risk of people unwittingly using the wrong calibration without receiving any kind of warning. Worse, some of the defaults are still R21 calibrations, so anyone currently using this is definitely getting something incorrect.

This MR removes the default configurations so the user must specify what calibration they're using. They can do this by providing configured calibration tools, or (new in this MR) by passing in the relevant config parameters to METSignificance, which will then set up the necessary tools with the correct calibrations. The former takes precedence if the user does both.

The only exception is for taus - here I have retained a default configuration, because analyzers do not have a free choice of which TES to use (it's fixed from AOD level) and they do not interact directly with this calibration. Hence they would be unlikely to know which calibration configuration to provide, as it's not part of analysis recommendations. The default value should be valid for all R22+ analysis for the foreseeable future. That said, users can still provide their own configuration if they want to.

Tagging @bhodkins, @stapiaar, @gotero for awareness. Thanks to @hpacey for noticing this issue in the first place.