Currently, the lab repair recommendations are selected along
these lines:
  * Make a list of DUTs that are nominally physically close
    together, and all in the same lab.
  * For each possible list, record the worst case "Avail"
    number for the lab after repairing all DUTs in the list.
  * Pick the list of DUTs that has the best possible worst
    case after repair.

However, selecting based on location frequently means selecting
DUTs of different boards.  It's preferable that we give priority
to repairing all DUTs of a given board instead.  So, we should
change the prioritization to be more along these lines:
  * Starting with the most important board (worst "Avail" number),
    and working down the list, select all DUTs for that board,
    until the number of DUTs to be repaired is at least the
    daily repair quota.
  * Repair lists should still be restricted to a single lab.