Avoid upssched infinite tight loops with (probably) closed sockets

[jimklimov]

Closes: #1964
Probably fallout from #1274 - CC @dtsecon for review

Now it is all quiet (literally, my fans whirred up without the fix):

:; killall upssched ; make -s -j 8 && \
    UPSNAME=heartbeat@localhost NOTIFYTYPE=ONBATT \
    NUT_STATEPATH=/dev/shm NUT_CONFPATH=`pwd`/tests/NIT/tmp/etc \
    ./clients/upssched -DDDDD

   0.000000     [D2] parse_at: is 'heartbeat@localhost' in AT command the 'heartbeat@localhost' we were launched to process?
   0.000052     [D1] parse_at: 'heartbeat@localhost' in AT command matched the 'heartbeat@localhost' UPSNAME we were launched to process
   0.000064     [D1] parse_at: processing CANCEL-TIMER
   0.000087     [D2] parse_at: is 'heartbeat@localhost' in AT command the 'heartbeat@localhost' we were launched to process?
   0.000094     [D1] parse_at: 'heartbeat@localhost' in AT command matched the 'heartbeat@localhost' UPSNAME we were launched to process
   0.000101     [D1] parse_at: processing START-TIMER
   0.000211     [D1] Keeping stderr open due to debug verbosity 5
   0.000257     Timer daemon started
   0.000265     [D2] Timer daemon waiting for connections on pipefd 10
   0.250397     [D3] new connection on fd 7
   0.250466     New timer: heartbeat-failure-timer (15 seconds)
   0.250513     [D1] Exiting upssched (CLI process)
   0.250586     [D4] read() from fd 7 returned 0 too many times in a row, aborting sock_read(); errno=0: Success
   0.250594     [D3] closing connection on fd 7

And that is it, the daemon is holding the clock.

  15.263748   Event: heartbeat-failure-timer
  30.311007     Timer queue empty, exiting
  30.311052     [D1] Timer queue empty, closing pipe and exiting upssched daemon

And it has truly exited:

:; ps -ef | grep -v grep | grep upssch

[jimklimov]

Hm, Jenkins says build #2 succeeded all 140 stages. So on coding side, PR is good to go, and does solve a practical problem.

On content side... gonna wait till next week if @dtsecon has time to comment about this change to his earlier fix in the area.

[jimklimov]

Rebased over master to stay clean, while waiting for reviews...