kstat(1): implement wait with setitimer(2) instead of nanosleep(2)

kstat(1): implement wait with setitimer(2) instead of nanosleep(2)

kstat(1)'s wait period drifts because nanosleep(2) uses a relative
timeout.  If we use setitimer(2)/sigsuspend(2) the period does not
drift.

While here, bump the upper bound for wait up to UINT_MAX and switch to
the normal strtonum(3) error message format.

With input from kn@.

Tweaked by bluhm@ to block SIGALRM with sigprocmask(2) while we're
outside of sigsuspend(2).

Thread: https://marc.info/?l=openbsd-tech&m=160038548111187&w=2

Earlier version ok millert@.

ok bluhm@