merging: account for sign in sqrtS fallback
Another one from the Sherpa validation suite: when we switched from beam energy to pz for the BEAMINFO
object, we should've accounted for the signs when using the momenta to work out sqrtS
in the fall-back calculation at merge time.