Pass igraph (per-event MLM graph) through Fortran bridge to SMATRIX_MULTI

[Copilot]

• Fix igraph1_page in process_sigmaKin_function.inc (CODEGEN only):
  • Renamed igraph1_page → igraph1_ievt, now indexed by allIgraph[ievt] (per-event) instead of allIgraph[ievt00] (first in page)
  • Moved the entire color-selection block inside the for( int ieppV ... ) loop so iconfig is determined per-event
  • Changed targetamp from fptype_sv (SIMD vector) to fptype (scalar), extracting lane [ieppV] via jamp2_sv[icolC][ieppV] in SIMD mode
  • Removed SIMD-only volatile (no longer needed with scalar targetamp)
  • For mixed-precision path: ievt2 now has its own independent per-event igraph read (igraph1_ievt2 = allIgraph[ievt2]), its own iconfig2, and its own targetamp2 scalar array — all independent of ievt

[Qubitol]

So, I have 2 things which I'd change:

• I would squash/rebase the commits, so that we have only one single commit making the changes inside CODEGEN (also, it seems it didn't remove all the changes done elsewhere, so I'd remove them). I'm saying that because for MadGraph7 I'd like to cherry-pick the commit, and having unrelated things makes it more annoying.
• we can do a second commit that updates the processes (but in that case we use the script to properly regenerate all of them)
• I would upgrade igraph to a class member (to be consistent with the rest), and not just a variable passed around

[oliviermattelaer]

No problem for squashing.

I would upgrade igraph to a class member (to be consistent with the rest), and not just a variable passed around

I have asked the agent to do that (he will upgrade the PR)

we can do a second commit that updates the processes (but in that case we use the script to properly regenerate all of them)

Let's wait for the final form and then tackle that

[Qubitol]

I think the core is done, but still there are changes in the various processes. I guess we can solve this by hand via interactive rebase.

[oliviermattelaer]

Ok I have asked the agent to remove the spurious change. Do we squash commit now?
or run the update script first? (or just squash now and then run the update script?)

[Qubitol]

CI passes! We can merge as soon as MadGraph v3.7.1 is released, bumping the version beforehand here.