topology: Use cpumasks to iterate all possible or online CPUs

[htejun]

CPU IDs aren't guaranteed to be consecutive. To iterate all possible or currently online CPUs, the respective cpumask should be iterated. There are several issues in this area:

• Some are iterating 0 to the number of online CPUS. This is incorrect as CPU IDs can have gaps and thus can miss higher numbered CPUs.
• Some are iterating 0 to nr_cpu_ids. This is correct but not ideal in that the CPU IDs can have substantial gaps not online or unallocated.
• The possible and online cpumasks can be accessed from BPF side using scx_bpf_get_possible_cpumask() and scx_bpf_get_online_cpumask(), respectively. However, scx_utils::toplogy doesn't expose the current possible or online CPU masks that are available through /sys/devices/system/cpu/possible and online.

Make scx_utils::topology expose the cpumasks and make all schedulers use the appropriate cpumask for iteration in both userspace and BPF code.

[uran0sH]

If no one works on it, I would like to work.