Reminder of a better way to discard helicity configs for loop diagrams.

In the first phase-spaces points considered for a process, NLOComp should decide which loops are consistently zero for some set of helicity configuration. The way this is implemented in version 81 is not optimal yet: SigVirt.f has a running variable MaxWgt which stores the maximum amplitude encountered so far when computing the loop amplitudes of a given phase-space point. If the amplitude from one loop is much smaller (with a threshold set by the user) than MaxWgt, and this consecutively on the first couple of phase-space point, then this helicity configuration is discarded for this specific loop.
An upgrade to this method which is safer in case of large number cancellation between loops (As it happens with triangle fermionic diagrams for example) is to compare the amplitude of one loop squared against the born with the total physical contribution for this phase-space point. The implementation is more involved of course, but it would insure to only discard contributions which really are negligible.