Use monotonic clock for monitoring about nut HOT 1 OPEN

@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

from nut.