Use monotonic clock for monitoring

@vencik: here is the result of my review, back in December, aborted due to issues with dummy-ups

iirc, this last was not satysfying the sleep delay, when using .seq file (check nut/data) That stopped my review
a minor remaining concern was that, with svn r3775, we are falling back to perl on Solaris, without being sure it's present... not a big deal since we're in the developer land, and have cascaded deps from autoconf, which requires perl. but still, that's a point that would be good to address.
tests were implemented / behaving as expected, apart from the dummy-ups issue, raised above. I've also executed some series of tests on the drivers and in general (including upssched), to ensure everything was still doing fine.
in common.c->vupslog(): I'm not a great fan of the start_init boolean method you've used. I would d have preferred something in the spirit of the previous implementation, i.e: ... start = { 0 }; ... if (start.FIELD == 0) // not init'ed ...nut_clock_copytime...

This would require to add a NONE value (set to zero) in nut_clock_mode_t

the riello driver, a new driver that has appeared after the MONOTONIC branch completion, needs update.
nut_clock.h is somehow redundant with timehead.h, though not addressing the same scope! But since timehead.h was in charge of including the right [sys/]time.h, which is now also addressed with nut_clock.h, there should be some possible simplification.

In the same spirit, some reference / update in the developers.txt doc (grep for "timehead.h") would be appreciated.

I've committed the "Net-SNMP includes mess" separately (r????).
You changed dstate-hal.h (dstate_poll_fds()), but not dstate-hal.c not a big deal since this code is dead (I'm keeping it for a potential evolution to upower). but I was pretty sure you knew how to use grep ;-p
few typo errors here and there

networkupstools / nut