more agressive pruning of diagram without flavor

[oliviermattelaer]

run the trimming

1. before color computation now
2. in all cases.
   By default this also remove diagram only possible for symmetric flavor that the one considered.
   the standalone --mask=False does not trim those last one for a simplify debugging/check

[theoheimel]

I tested the two processes from #13 (comment) and both run now and the cross section in MG7 appears to be correct. So from my side, we can merge this

[oliviermattelaer]

I would say that i would like to check the following:

• why the build wheels is suddently crashing (but maybe we do not care --your call--)
• MG7 Hall of Shame #13 (comment) might be created by this patch, so I will need to investigate this (first crash is fixed and does not seem related but another related one surfaced later, and I did not had the time to fix it yet)

[theoheimel]

The build wheel issue was unrelated and I fixed it in main

[oliviermattelaer]

update on the status for w+ _anti_neutrino > z ta+

• run does generate/output
• check language is consistent

Language comparison results (Fortran SA / C++ SA / MG7 SA / Python):
Process          Fortran SA      C++ SA          Python          max rel.diff.   Result
w+ vt~ > z ta+   2.045462e-01    0.000000e+00    0.000000e+00    1.000000e+00    Failed

• check is working:

Lorentz invariance results:
Process          Min element       Max element       Relative diff.    Result
Summary: 0/0 passed, 0/0 failed
Not checked processes: w+ Nx > z ta+
Flavor grouping check results:
Process          Merged            Unmerged          Relative diff.    Result
w+ vt~ > z ta+   0.0000000000e+00  3.7927733170e-01  1.0000000000e+00  Failed
Summary: 0/1 passed, 1/1 failed
Failed processes: w+ vt~ > z ta+
Gauge results (switching between Unitary/Feynman gauge):
Process          Unitary           Feynman           Relative diff.    Result
Summary: 0/0 passed, 0/0 failed
Process permutation results:
Process          Min element       Max element       Relative diff.    Result
w+ Nx > z ta+    0.0000000000e+00  0.0000000000e+00  0.0000000000e+00  Passed

Note "manual" check of merged/unmerged (with fortran output) are consistent .... So the issue is not fortran related but purely C++ and Python that does return zero

[oliviermattelaer]

update on the status for w+ _anti_neutrino > z ta+

• run does generate/output
• check language is consistent

Language comparison results (Fortran SA / C++ SA / MG7 SA / Python):
Process          Fortran SA      C++ SA          MG7 SA (cppnone)MG7 SA (cppsse4)Python          max rel.diff.   Result
w+ vt~ > z ta+   2.045462e-01    2.045462e-01    2.045462e-01    2.045462e-01    2.045462e-01    2.156695e-08    Passed
Summary: 1/1 passed, 0/1 failed

• check is working:

Lorentz invariance results:
Process          Min element       Max element       Relative diff.    Result
Summary: 0/0 passed, 0/0 failed
Not checked processes: w+ Nx > z ta+
Flavor grouping check results:
Process          Merged            Unmerged          Relative diff.    Result
w+ vt~ > z ta+   2.3615583289e+01  2.3615583289e+01  0.0000000000e+00  Passed
Summary: 1/1 passed, 0/1 failed
Gauge results (switching between Unitary/Feynman gauge):
Process          Unitary           Feynman           Relative diff.    Result
Summary: 0/0 passed, 0/0 failed
Process permutation results:
Process          Min element       Max element       Relative diff.    Result
w+ Nx > z ta+    0.0000000000e+00  0.0000000000e+00  0.0000000000e+00  Passed
Summary: 1/1 passed, 0/1 failed

@Qubitol, check carefully the list commit since it change some deep stuff for the madmatrix side (which was creating some segfault for this process)