If a CPU is marked disabled, we should not create a CPU device for it.
This is present in some RISC-V device trees which contain a "monitor
core" with limited functionality (no MMU). We don't run the kernel
on these, and in early CPU start-up code we skip them, and they have no
impact on mp_ncpu. Why this is included in the device-tree at all is
beyond me. It seems the new trend is to mark these as "disabled", so
let's respect this in the ofw_cpu probe routine.
This is generally harmless, but I noticed an impact when attempting to
attach the cpufreq_dt driver. It counted more OFW CPU devices than
logical CPUs (mp_ncpus), and therefore failed to attach for logical CPU
3 (RISC-V hart 4). With this change, all is well, and we don't get a
bogus device.
No functional change when the property is not present at all.