Add poorman's trigger training outputs

This is mostly changes to study a "poorman's trigger" which only considers the impact parameter with respect to the beam spot, i.e. no primary vertex or secondary vertices at all. I suspect we might actually be able to do pretty well. This would be awesome in the trigger because reconstructing enough tracks to build a primary vertex is real expensive.

It touches a bit of common code, though, so maybe better to get a quick look from @alfroch or @mguth. The changes were:

• Rename output_remapping configuration field -> remapping since it does a lot of input remapping too. This was changed in all the config files but should have no effect on the outputs.
• Put d0Uncertainty and z0SinThetaUncertainty in the list of custom track info dumpers. These are not called in the offline code, so nothing should change. We already have the option to take d0 and z0SinTheta from FlavorTagDiscriminants, in place of the btagIp_* variables. In trigger I'm switching to them to be more generic: if we change the impact parameter definition the old btagIp_* variables would have to be renamed too, and it seemed like it would be easier to rename the track collection in Umami than several variables.

I also created another "fragment" called trigger-track-variables.json. Here I'm using the variables I added above. Again, I didn't want to change the name of the track outputs in code used by the offline group, but it might be worth moving cohesively to d0 and z0SinTheta rather than btagIp_d0 and btagIp_z0SinTheta. If the offline group wants that I can make another MR, which will put everyone on the same track variables.