libpfc timer should include kernel cycles too

[travisdowns]

Currently we are only getting user-mode cycles from libpfc. We should probably track all cycles.

[travisdowns]

Compare for example:

./uarch-bench.sh --test-name=syscall/* --timer=libpfc --extra-events=skl::UNHALTED_CORE_CYCLES

which gives:

** Running group syscall : Syscall benches **
                               Benchmark    Cycles    UNHALT
                     getuid() glibc call     48.00    706.69
                  getuid using syscall()     53.00    706.41
                  getpid using syscall()     53.00    726.17
            close() on a non-existent FD     69.00    729.68
                    non-existent syscall     72.00    665.44
                   getuid direct syscall     41.00    699.77
             non-existent direct syscall     41.00    631.84
                   syscall+lfence before     51.36    641.31
                    syscall+lfence after     52.32    644.63
                     back-to-back lfence      4.88      4.88
                         parallel-misses     16.93     16.96
                          mfenced misses    234.90    235.05
                          sfenced misses     17.18     17.20
                          lfenced misses    181.29    181.40
                          syscall misses    234.79    831.68
               misses + lfence + syscall    244.98    835.93

Note how the Cycles column is impossibly low, while the skl::UNHALTED_CORE_CYCLES (UNHALT) column shows the correct time.