wbsio_detach never detaches its lm child (no bus_generic_detach): driver unload leaves a dangling attached child with a live 5s sensor task, leaked rman ports, and a NULL-ops parent — any later bus op on the child is a guaranteed NULL-deref panic

Summary

The detach violates the newbus bus-driver contract - devclass_delete_driver skips grandchildren, so the lm child stays DS_ATTACHED with a live sensor task while the parent's softc is freed and kobj ops NULLed; the next BUS_* dispatch through the parent (e.g. kldunload lm) KOBJOPLOOKUPs NULL -> guaranteed panic (UAF in module builds). Root-gated module-build + Winbond-hardware gated (probe reads 0xff on QEMU - untestable). bus_generic_detach one-liner fix.