fix(hw): allowlist stock-Kuiper ZynqMP peripheral probe failures

The new ``assert_no_probe_errors`` check fires on three Kuiper
ZCU102 boots that have nothing to do with the ADC/DAC/JESD flow
under test:

- ``cdns-wdt: probe of ffcb0000.watchdog failed with error -2``
- ``zynqmp-display: probe of fd4a0000.display failed with error -22``
- ``ahci-ceva: probe of fd0c0000.ahci failed with error -22``

All three fire on every Kuiper ZCU102 boot regardless of the
overlay — the hardware is either unconfigured (DisplayPort with no
monitor attached), the clock tree isn't routed (Cadence WDT), or
the peripheral isn't wired out on the dev board at all (Ceva AHCI
/ SATA).  Match by the specific device-node address so a genuine
regression on the same driver type (e.g. a real watchdog failure
elsewhere) still trips.

Author: tfcollins

test/hw/hw_helpers.py

@@ -308,6 +308,14 @@ def require_hw_prereqs() -> None:
     "irqchip_init",
     "__primary_switched",
     "rpu-bus/interrupt-controller",
+    # Stock Kuiper ZynqMP (ZCU102) probes these SoC peripherals from the
+    # base DTS regardless of the overlay we merge in — the hardware is
+    # either unconfigured (no DisplayPort monitor attached) or not wired
+    # out on the board (no SATA).  Match by device-node address so a
+    # genuine regression on the same driver elsewhere still trips.
+    "ffcb0000.watchdog",  # Cadence WDT — unroutable clocks
+    "fd4a0000.display",   # ZynqMP DisplayPort — no monitor + DPMS pipe
+    "fd0c0000.ahci",      # Ceva AHCI/SATA — not routed on ZCU102
 )
 
 # Hard-fail patterns — these indicate a genuine kernel fault.