Remove SIGTERM handler override in AM gather and send plugins

[robjarawan]

What

gather/am.py and send/am.py both override SIGTERM with SIG_DFL in __init__. This kills the process immediately on sr3 stop/restart, bypassing the flow's graceful shutdown — on_stop callbacks, message acking, and connection cleanup are all skipped. Data loss on every stop/restart cycle.

Change

Remove the signal.signal(signal.SIGTERM, signal.SIG_DFL) call from both files so the parent flow's handler runs properly. Removed unused signal import.

[petersilva]

I'm puzzled about how to test this... I'm not sure it has the effect hoped for. would want some testing.

[robjarawan]

Fair point. The change is small -- just removing the SIG_DFL override so the parent flow's handler runs instead. The expected effect is that cleanup (closing sockets, stopping loops) actually happens on SIGTERM instead of an abrupt exit. I don't have an AM setup to test against but I can verify that the parent signal handler fires correctly with a mock. Would that be enough, or do you want to see it tested against an actual AM feed?

[petersilva]

Maybe @andreleblanc11 has some configs he can test this branch with? you need some AM flows, right?

[petersilva]

marking low priority because AM traffic is (gradually) being de-commissioned. so... yeah fine to fix it, but not super urgent.

[andreleblanc11]

We added the signal handler to the AM code before Reid added it to the sr.py in https://github.com/MetPX/sarracenia/pull/559/changes.

So I think it's probably OK to remove but will need to test.

[robjarawan]

The diff itself is strictly safer: it only deletes the two SIG_DFL overrides plus imports, restoring the parent flow's graceful handler with zero logic change.

One merge condition: a mock-handler smoke test, since bare fixtures do not construct (Am needs sendTo, fileSizeMax, add_option, and a stubbed socket). I verified the corrected test fails pre-PR on both modules and passes post-PR. Happy to push it on request.

The deeper items I found, the infinite accept loop with no stop check and the fork/execl orphan window, are pre-existing and I am marking them wontfix. AM is being decommissioned, so no loop rework is justified for a 6-line cleanup. @andreleblanc11 a live-flow eyeball on your side would still be welcome before merge.