test/integration: wait for startup before timing, keep failures visible

Review fixes. The first-boot systemd-analyze now waits (bounded) for
startup to finish, since SSH can answer first and systemd-analyze
refuses until it has. In the boot scenario, a non-zero exit from
is-system-running (as for "degraded") no longer ends the script under
set -e before the diagnostics are saved, and a failed query for failed
units writes a marker instead of an empty file that read as "none".

test/integration.d/base-test.sh

@@ -574,7 +574,10 @@ install_phase() {
   # without attaching the disk.
   ssh_sudo "cat /var/log/omarchy-install-timing.json" >"$BASE_DIR/omarchy-install-timing.json" 2>/dev/null || true
   ssh_sudo "cat /var/log/omarchy-install.log" >"$BASE_DIR/omarchy-install.log" 2>/dev/null || true
-  ssh_guest "systemd-analyze" >"$BASE_DIR/first-boot-systemd-analyze.txt" 2>/dev/null || true
+  # SSH can answer before startup finishes, and systemd-analyze refuses until
+  # it has; wait (bounded) so the first-boot numbers are final.
+  ssh_guest "timeout 120 systemctl is-system-running --wait >/dev/null; systemd-analyze" \
+    >"$BASE_DIR/first-boot-systemd-analyze.txt" 2>/dev/null || true
 
   log "Installed system is up. Saving base image."
   stop_vm

test/integration.d/boot-test.sh

@@ -19,11 +19,15 @@ boot_ms=$((($(date +%s%N) - boot_started) / 1000000))
 log "SSH answered within ${boot_ms} ms of power-on"
 
 # --wait returns once startup has finished, so systemd-analyze has final
-# numbers and a unit that fails late in boot is still counted.
-state=$(ssh_guest "timeout 120 systemctl is-system-running --wait" 2>/dev/null | tr -d '\r')
+# numbers and a unit that fails late in boot is still counted. It exits
+# non-zero for "degraded"; keep the answer so the diagnostics below are still
+# collected and the assertion reports it.
+state=$(ssh_guest "timeout 120 systemctl is-system-running --wait" 2>/dev/null | tr -d '\r' || true)
 ssh_guest "systemd-analyze" >"$RUN_DIR/systemd-analyze.txt" 2>/dev/null || true
 ssh_guest "systemd-analyze blame --no-pager | head -n 20" >"$RUN_DIR/systemd-analyze-blame.txt" 2>/dev/null || true
-ssh_guest "systemctl --failed --no-legend --plain" >"$RUN_DIR/failed-units.txt" 2>/dev/null || true
+# A query that fails must not read as "no failed units".
+ssh_guest "systemctl --failed --no-legend --plain" >"$RUN_DIR/failed-units.txt" 2>/dev/null ||
+  echo "could not query failed units over SSH" >"$RUN_DIR/failed-units.txt"
 printf '{"boot_to_ssh_ms": %d, "system_state": "%s"}\n' "$boot_ms" "$state" >"$RUN_DIR/boot-timing.json"
 
 check "systemd reports the system running (got: ${state:-nothing})" test "$state" = running