x86 gate: simple_exit's by-design exit 42 is counted as a failed process, so run-boot-parallel.sh fails on main

[ryanbreen]

./docker/qemu/run-boot-parallel.sh 1 fails on main because simple_exit
exits 42, which is the only thing that program does.

What happens

[ INFO] kernel::syscall::handlers: TEST_TALLY: exited=22 nonzero=1 failed=[simple_exit:42]
[ERROR] kernel::syscall::handlers: TEST RUNNER: FAILED - 1 of 22 userspace processes exited nonzero
x86 userspace gate: FAIL - failing process is not allowlisted: simple_exit
Results: 0 passed, 1 failed out of 1

The boot itself is healthy: it reaches the
USERSPACE TEST COMPLETE marker first, and STRAND_CENSUS reports
threads_saved_blocked=11 stranded=0.

Why it is not a kernel failure

userspace/programs/src/simple_exit.rs is 3 lines of body:

fn main() {
    // Just exit with code 42 - no printing
    std::process::exit(42);
}

Its doc comment says the same: "Just exits with code 42, no other output."
Exiting 42 is the behaviour under test, so the tally's "nonzero exit means a
failed process" rule mislabels it, and
scripts/x86-gate-verdict.sh then fails the run because
scripts/x86-gate-allowlist.txt holds 0 entries (3 comment lines and nothing
else).

Where it reproduces

Measured on beast, breenix-x86, /root/breenix-a64r2, one run each, each
after touch kernel/src/task/scheduler.rs and a
--features testing,external_test_bins rebuild:

Revision	Result
52491c4b (main, 2026-09-02)	Results: 1 passed, 0 failed out of 1
bfbb7575 (main, 2026-09-04)	TEST_TALLY: exited=22 nonzero=1 failed=[simple_exit:42], FAIL
d7679a7b (fix/562-761-aarch64-testing-profile round 2)	same tally, FAIL
8a52c22f (same branch, round 3)	same tally, FAIL, 2 of 2 runs

So it arrives on main between 52491c4b and bfbb7575, and the aarch64
testing-profile branch inherits it rather than causing it. 1 of the 1 run at
the older main passed, so the gate was green there.

What would close it

Either the tally stops counting a program whose contract is a nonzero exit as
a failed process, or simple_exit's expected code is declared somewhere the
tally reads. Adding it to scripts/x86-gate-allowlist.txt would be the wrong repair: that
file's own header says a listed test must still fail, and this test is not
failing.