arm_scmi: Fix infinite recursion in transport probe

[USAVault]

Hi CIX-ians, thanks for your hard work! :) I'm trying to run your kernel by applying patches to my OS, and ran into an issue

When DEFINE_SCMI_TRANSPORT_DRIVER() creates an "arm-scmi" child device for each transport (SMC, mailbox, etc.), it also copies the parent's dev->of_node along with the rest of the fwnode. So the child ends up with the same OF node as its parent
When the child device is added, platform_match() calls of_driver_match_device(), sees the node matches the transport driver's of_match_table, and probes away spawning another "arm-scmi" -> another probe -> infinite recursion -> stack overflow
To break the cycle, keep the fwnode, but use device_set_of_node_from_dev() for of_node instead. This sets of_node_reused = true, so of_match_device() skips it and the transport driver never matches its own child.
Bonus: refcounting is now balanced too,, device_set_of_node_from_dev() does of_node_get(), which pairs with of_node_put() in platform_device_release()

[amazingfate]

Thank you about this. We are focusing on acpi now so this issue is missed. It's nice to fix it.