fsm.c is the only user of timed_printf, and actually I don't expect
other messages to need timing (since we know the time from the fsm
states).

I may repent, I'm aware. But at this point I'm killing diag-yes, to
make the one and true diagnostic environment, without duplications and
conflicts.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>