Convert consumers to use the CPU_FOREACH_SET() macro

[markj]

Summary

This implementation is faster and doesn't modify the cpuset, so it lets
us avoid some unnecessary copying. No functional change intended.

[kib]

BTW there is enough similar iterations FFS/CLR done over the signal sets/masks. Do you plan to look at this as well?

[markj]

In D32029#722803, @kib wrote:

BTW there is enough similar iterations FFS/CLR done over the signal sets/masks. Do you plan to look at this as well?

I will try implementing this for sigsets too. I guess it would be somewhat messy to change the kernel to use bitset(9) for signal sets?

[rlibby]

sys/amd64/vmm/io/vlapic.c

895–897

This is currently broken due to the issue with break noted in D32028.

sys/amd64/vmm/vmm_lapic.c

92–94

Again, break is broken.

[markj]

sys/amd64/vmm/io/vlapic.c

895–897

Thanks, I reverted this commit for now.