Thank you Bradley!  Quite well done, both in terms of style and
completeness; I particular appreciate your adding both ATF and
system tests.  Things I changed during review:

- the command syntax "yyyymmddvv" seemed likely to be hard to 
  remember, so I just replaced it with "date".

- the unit tests weren't quite effective as written: the "old" serial
  number was being set incorrectly as a unixtime value, so the "new"
  serial never came into conflict with it.  changing to, for instance:

      /* future to date */
      set_mystdtime(2014, 4, 2);
      old = dns_update_soaserial(0, dns_updatemethod_date);
      set_mystdtime(2014, 4, 1);

  ... enables the test to correctly check the behavior of a serial
  number going forward or backward in time.

- the julian date converter you used in epoch_to_yyyymmdd() is nifty,
  but a) I'm not capable of evaluating its correctness and b) it's
  inconsistent with the way we've done similar conversions elsewhere
  in BIND, so I switched it to use gmtime().

- there was no documentation; I added a paragraph to
  doc/arm/Bv9ARM-book.xml